How Microsoft Fara1.5 Local Multimodal Web Agent Navigates

Introduction

The automation of digital processes was always hampered by the vulnerability of application metadata. The software frameworks that aimed at automating web navigation were always hindered by the inherent instability of the source code. Any change in the website structure immediately broke the old automated scraping scripts.

The innovative approach takes care of this problem by presenting a vision-based online orchestrator designed to handle the actual operations, including multi-step inputs and various catalog products. By eliminating any form of dependency upon the source structure and running its processes inside an isolated runtime environment, the proposed solution creates an effective scaling framework within the user machine. Enterprises can run their highly precise workflows locally without sending unencrypted visual interfaces to huge cloud server clusters and dealing with the additional API layer. The new paradigm series is known as Fara1.5.

What is Fara1.5?

Fara1.5 is family of vision-only browser automation models developed by Microsoft Research to serve as highly efficient computer-use agents. Built upon a multimodal decoder-only structure fine-tuned from a Qwen 3.5 base architecture, the model family interacts with software applications exclusively by analyzing raw user interface screenshots and emitting structured tool actions. By completely bypassing traditional document object models (DOM) and accessibility tree paths, Fara1.5 operates visually, matching or outclassing the capabilities of massive proprietary cloud models while remaining small enough to run locally within a sandboxed, virtualized environment.

Model Variants

• Fara1.5-4B : The smaller 4B version is designed to work on edge scales and therefore provides an effective runner locally for consumer devices without having to invest in costly cloud-based computing resources. This version works effectively to show that small models are capable of achieving very high levels of completion of tasks in live-web tests without exposing any local variables or files of corporate nature to the data servers.
• Fara1.5-9B : As the name suggests, this version is the centerpiece of the entire family of models and should be used by most enterprises in their automation tasks. It is based on the '2/3rds Rule' of scalability, which implies that it achieves two-thirds of the efficiencies that come from full scaling of the version from 4B to 27B. It is thus an excellent model for compute efficiency and reasoning. In addition, it doubles the success rate of 7B models with a bigger 262K context window.
• Fara1.5-27B : The Fara1.5-27B model belongs to the highest performing version of this set, designed explicitly for achieving the highest levels of execution performance in highly nested websites. The top model introduces cutting-edge performance standards for the pixel-to-action models, which are designed precisely to take care of advanced cross-site transactional tracking along with massive information gathering capabilities, which normally exceed the scope of generic models.

Key Characteristics of Fara1.5

The fundamental strengths of Fara1.5 are derived from a collection of intrinsic features that distinguish it from generic prompt iteration systems and earlier automated systems:

• Absolute Coordinate Prediction: Instead of depending on external cues or the set-of-marks system, which fails at higher resolutions of the application's display interface, Fara1.5 has the ability to determine absolute spatial coordinates.
• Active Context Management Actions: Possessing a context window of 262K tokens, the system makes use of a special action called Memorize. It ensures that the system actively keeps track of the essential details, such as comparing the price on different vendor webpages, thus preventing hallucinations that can happen if the pertinent information moves out of the field of view.
• Ambiguity Resolution with Operator Collaboration: As opposed to generic automated agents that follow an 'autonomy or failure' principle of operation, Fara1.5 is trained to prompt operators with questions when faced with ambiguous instructions by the user.
• Baked-in Critical Point Protocol: To mitigate financial and operational risk, the underlying training protocol of the model incorporates an unequivocal safety rule when it comes to state-changing and non-reversible decisions. At a point where there is critical decision making—such as clicking on a buy-now button, signing up a contract, or entering a personal identifier—the program prompts for a human go-ahead.

Use Cases of Fara1.5

• Privacy-Preserving On-Device Field Agency : In environments where there is significant corporate regulation and compliance-mandated restriction of data movement, the small-sized 4B model may be run natively on the device used by the employees themselves. This would be useful for agents helping employees complete forms and verification processes regarding internal audits or HR records. Since the agent will run on-device, the context of any private individual data or screenshots of internal corporate workings will remain within the confines of the machine's memory.
• Cross-Platform Identity and Context Syncing : The well-rounded 9B model may be used as a context orchestrator, capable of fluid switching between multiple programs which require secure log-in. By using its contextual and memory capabilities, the agent will be able to log into the program's interface, determine the required software information, open up a second program that holds a calendar, and synchronize projects with complete semantic coherence across two applications.
• High-Risk Transactional Bulk Audit : For companies that manage huge logistics operations, the leading 27B model can be employed for conducting automated bulk comparison shopping and contract auditing. The 27B model is able to handle multiple interfaces at once in order to make sure that the current prices correspond to contractual agreement. With its own critical points safety protocol, it makes sure that in case of any discrepancy such as an abnormal price drop or an ambiguous invoice calculation, it will immediately stop in order to seek human intervention before automatically conducting a transaction worth thousands of dollars.
• Interoperability Layers for Legacy Web Software: For companies using old-fashioned proprietary software without APIs, the entire set of models from Fara1.5 can serve as a universal interoperability layer. Due to the fact that the model understands interfaces only via screenshot, it can work with very old interfaces with unmapped interactive objects and complicated forms. This way, developers can easily automate workflows on legacy software without reconstructing broken accessibility trees or noisy DOMs.

How Does Fara1.5 Work?

The key to understanding the functioning of Fara1.5 lies in its gradual approach to planning that operates within an extremely concise observe-think-act feedback loop. The exact procedure that goes into making Fara1.5 function is outlined in the workflow flowchart given below:

source - https://www.microsoft.com/en-us/research/articles/fara1-5-computer-use-agent/

1.Context Capture:(Step 1) – The model takes in the initial textual instruction from the user, the action history log, and precisely three latest screenshots from the browser.

2.Internal Cognitive Processing:(Step 2) – Fara1.5 processes the visual context using its multimodal decoder-only model architecture to extract spatial coordinate matrices and correlate data points with factual information stored internally by the model.

3.Ambiguity and Safety Checks:(Step 3) – Internal safety modules perform safety checks on the action path suggested by the model. In case the current action corresponds to any of the critical checkpoints with ambiguity in instructions, an intervention flag is raised.

4.Structured Tool Output:(Step 4) – After the successful completion of safety checks, the model generates a single action tool output (e.g., click, type, scroll, web_search, and visit_url) based on the training loss only for the latest turns. 

The key component responsible for enabling Fara1.5's sophisticated functionality is the FaraGen1.5 and FaraGen2.0 training procedures developed by Microsoft. This multi-agent system uses a highly capable GPT-5.4 teacher solver that creates millions of high-quality synthetic browser paths. To prevent the student models from learning how to navigate through algorithmic tricks, the teacher solver is not allowed to perform any URL query-based manipulation in order to reach the destination web page.

How Fara1.5 Learns?

Apart from that, when dealing with concerns regarding the presence of poor-quality data, due to the need for safe user login in gateable regions, the use of programming languages has been seen in code tools like GitHub Copilot CLI, for creating sandboxed local clones of popular websites for emails, calendars, and management, called FaraEnvs, which help in training the model for real user logins. Data is evaluated according to its quality through an automated gating system that evaluates each trajectory on the basis of three factors: correctness (through a high-powered privileged-information LLM judge that verifies each state change by assessing the difference between the database snapshots pre-task and post-task), efficiency (by punishing redundant mouse clicks), and safety (ensuring that the model pauses at appropriate junctures for user decisions).

source - https://www.microsoft.com/en-us/research/articles/fara1-5-computer-use-agent/

High-quality semantic coherence between applications has been ensured by using FaraGen1.5 for creating persona-consistent narratives (IT company worker personas, in this case) while operating with different applications. Contextual noise has been managed effectively through selecting only the most salient screenshots from a series of shots for validation purposes.

Performance Evaluation with Other Models

In an evaluation using the Online-Mind2Web benchmark, which consists of 300 highly complex tasks divided across 136 live, unsandboxed webpages, the Fara1.5 models showcase clear superiority over open-weight baselines and huge closed-source proprietary systems. The main Fara1.5-27B variant establishes itself as a new benchmark for pixel-to-action models thanks to a superior 72.0% task success rate, giving it a whopping +13.7% performance advantage over the OpenAI Operator with its 58.3% success rate on the same testbed. From the comparison metrics, the high performance density of the small open weights is obvious as the relatively balanced Fara1.5-9B attains a task success rate of 63.4%, beating the second-best open baseline GUI-Owl-1.5-8B's score of 48.6% while equaling that of the closed system such as the Yutori Navigator n1 with 64.7% success rate. Not even the edge Fara1.5-4B fails to impress as it attains a decent task success rate of 57.3%, matching Google's far bigger Gemini 2.5 Computer Use model's capability.

source - https://www.microsoft.com/en-us/research/articles/fara1-5-computer-use-agent/

Outside the conventional web browsing assessment, other benchmark tests validate the superiority of the family with respect to stability and consistency. In the case of visual navigation assessment through the WebVoyager benchmark test, Fara1.5-27B achieves an advanced accuracy rate of 88.6% compared to the 87.0% achieved by OpenAI Operator. In addition, similar performance is recorded in long-tail enterprise tasks in the WebTailBench v1.5, where 9B model performs +8.2 better than 7B model.

How to Access and Use Fara1.5?

Fara1.5 is a publicly accessible open-weight version available through the Microsoft Foundry platform. While the 9B version of this system is already active at present, the 4B/27B versions will be coming up soon. The best way for engineers to deploy Fara1.5 locally is by using the official MagenticLite inference harness from the GitHub platform. This harness has to run strictly inside a dockerized environment.

Limitations and Future Work

The limitations of Fara1.5 only include interfaces that are able to speak English. Additionally, due to the way that sandboxes work, there are still ways for adversaries to use network access to attempt to insert harmful code using web page layouts as cover will pose as a major risk to the overall performance of the agent in the future. Future versions of Fara1.5 will have a wider range of uses for synthetic training across a wider range of applications and more visually diverse reasoning patterns.

Conclusion

By using a separation of the orchestration of abstract reasoning from the execution of the tool at the pixel level and hosting both locally within the hardware of the machine, Fara1.5 provides an alternative solution to traditional cloud-based solutions for the automation of tasks that has a high degree of security and reliability. The primary contribution of the Fara1.5 architecture is demonstrating that local sovereignty of data does not need to be negatively impacted by the ability to perform tasks well.

Sources:
Blog: https://www.microsoft.com/en-us/research/articles/fara1-5-computer-use-agent/
9B Model: https://ai.azure.com/catalog/models/Fara1.5-9B



Disclaimer - This article is intended purely for informational purposes. It is not sponsored or endorsed by any company or organization, nor does it serve as an advertisement or promotion for any product or service. All information presented is based on publicly available resources and is subject to change. Readers are encouraged to conduct their own research and due diligence.

Qwen 3.7-Max: 35-Hour Multi-Agent Workflows Without Human Input

Introduction 

Construction of platforms that can undertake independent actions calls for a paradigm shift away from traditional paradigms of prompt and response. Workflow automation in this sense is beyond simplistic text generation since it is very dependent on the development maturity of synthetic simulation ecosystems and dynamic training environments. In terms of system deployment, the primary concern has evolved into that of constructing tight couplings between the scaffold layers and the underlying computational architectures. This requires an extremely dense logical reasoning process along with scaffold dependability, enabling the systems to work through huge time scales without compromising on their functionality. 

In this environment, the New AI model emerges as a unique infrastructure solution. Through the extraction of intelligent information from diverse runtime platforms and not from a fixed set of textual databases, this system avoids the rigid format that often leads to failures within the automation process. The New AI model is an efficient solution for cases where tool manipulation and feedback are necessary on a continuous basis and reliable execution pathways are required throughout lengthy system timelines. 

What is Qwen 3.7-Max?

Qwen 3.7-Max is an internally developed proprietary model by Alibaba Cloud which acts as the base for building agents, as it has been developed for the specific purpose of working like an agent and handling all its functions. The reasoning capacity of Qwen 3.7-Max can stretch for very long distances; it comes equipped with its own internal verification process-based reasoning mode.

Key Features of Qwen 3.7-Max

Several architectural features are built into the model to ensure stability throughout long computations:

• Increased time horizon: Created with the purpose of stabilizing both the internal state and policies of the model during consecutive runs conducted without human input for up to 35 hours and involving more than 1,000 tool calls.
• Instruction and Context Robustness: The model is endowed with innate instruction resistance and robustness to context decay, allowing it to perform long-horizon computations that involve more than a thousand steps without forgetting its key goals
• Context Intrinsic Preserving: Has capabilities for the preservation of thinking to retain entire reasoning chains across several moves, preserving its decision-making logic at a deeper level and saving tokens in the process.
• Format-Invariant Flexible Tool Use: Unrestricted by structural interdependence, the model has achieved format-invariant tool use behavior that allows it to operate flexibly and logically despite changes in the environment's format or harness.

Use Cases of Qwen3.7-Max

• Multi-Horizon Project Condensation : Major projects such as comprehensive database reworking, predictive analytics modeling, and regulatory reports take about one or two weeks for engineering teams. By leveraging its capability of running for up to 35 hours continuously, the model condenses all these activities to take place in just one session. The model becomes an automated orchestrator that goes through code bases, generates migration scripts, runs tests for error detection, and documents the entire system for publication in one un-interrupted execution cycle.
• Strategic Risk Assessment & Simulation : For critical decision making processes, the model can generate thousands of market simulations for any turn horizon range. In times when the system is under operational pressure, it becomes a seasoned operator that autonomously identifies hidden risks, detects any fraudulent behavior in transactions, and bans risky client behaviors to concentrate on steady income streams.
• Autonomous Optimization for ‘Day-Zero’ Unseen Hardware : Traditional code generation requires thorough documentation of hardware and pre-compilation of software libraries to generate optimized code. However, Qwen 3.7-Max does not rely on such documentation and uses a robust in-context generalization mechanism. By being dropped into an undocumented hardware architecture such as that seen in customized silicon accelerators and even novel tape-outs including the T-Head ZW-M890 PPU, the model takes advantage of real-time compilation and profiling to write GPU kernels iteratively to obtain optimal hardware optimization.
• Self-Monitoring Watchdogs for RL Pipelines : Training large scale distributed systems via reinforcement learning often leads to training instability due to ‘reward hacking,’ where the machine learning model exploits vulnerabilities in the simulation environment and violates design constraints. Using Qwen 3.7-Max as an autonomous validation watchdog in live training loops would enable the detection of reward hacking by adversarially generating and introducing new heuristics in the environment.
• Long-Duration Physical Embodied Intelligence: Not only does the model transcend the traditional digital terminal command approach by integrating itself into physical execution through robotics-specific toolkits such as Qwen-RobotClaw and Qwen-RobotNav, but it also enables itself to be used as the core planning agent for such physical agents as robotic dog quadrupeds working in inspection areas or even search-and-rescue scenarios. Utilizing the long-duration physical interaction memory layer lasting up to 20 minutes, it is able to ensure constant and long-term planning without falling back on the sporadic frame-by-frame reactions found in normal multimodal visual models.

How Does Qwen 3.7-Max Work?

The key to the intelligence of Qwen 3.7-Max is the ability of the model to scale through an environment strategy that focuses less on memorizing benchmark information and more on problem-solving experience. The RL framework of this model uses a decoupled structure where training instances are divided into three independent elements: {Training Instance = {Task, Harness, Verifier}}. With cross-harness and cross-verifier RL scheduling processes, the model is prevented from developing training hacks and exploiting any biases of its environment, and therefore is trained to develop logic-based general solutions.

In order to ensure policy consistency through long periods of time during training, tasks themselves are formulated as cumulative survival games that grow increasingly complex with each new training instance. Such scaling of temporal complexity ensures the penalty for committing early logical mistakes that could result in failures later during the trace. The model learns to perform continuous self-verification, allowing it to perform multi-hour-long, branched operations with no sign of cognitive fatigue.

Performance Evaluation with Other Models

When it comes to the main performance evaluation of the autonomous agent behavior, Qwen 3.7-Max manages to prove its superiority in the Terminal Bench 2.0 tests. According to Table below, the model managed to get the highest score of 69.7, easily beating DeepSeek-V4-Pro Max (67.9) and its previous version, Qwen 3.6-Plus (61.6). Moreover, it obtained 60.6 points on the SWE-Pro coding repository task and competes fiercely with the Claude Opus family. This evaluation is vital for engineering tasks since it confirms the ability of the model to work in unattended terminals, perform multi-step commands, and debug codes independently.

source - https://qwen.ai/blog?id=qwen3.7

The second important evaluation centers on the ability of the model to manage multi-agent workflow through the MCP-Mark (Protocol Agility) benchmark test. According to table above, the Qwen 3.7-Max scored impressively by scoring 60.8, decisively placing it ahead of GLM-5.1 (57.5). When put into perspective, it should be stated that the intelligent system succeeded in solving the extremely challenging GPQA Diamond test of logical reasoning with a score of 92.4, surpassing Claude Opus 4.6 (91.3). 

source - https://qwen.ai/blog?id=qwen3.7

The importance of the evaluation in terms of enterprise productivity cannot be overstated since the model is proved capable of functioning as a robust backbone for orchestrating office automation perfectly. In the business simulations such as YC-Bench, the system made $2.08M in revenues for a company, nearly doubling the performance of its direct predecessor, Qwen 3.6-Plus, which achieved $1.05M.

How to Access and Use Qwen 3.7-Max? 

The service is provided as a paid, proprietary model available on the Alibaba Cloud Model Studio API. Designed to integrate seamlessly into the current architecture of enterprises, the model is fully compliant with OpenAI/Anthropic APIs and request format standards. The model can be employed as a backbone within the top-tier production agent software such as Claude Code without changing any orchestration logic.

Limitations

While Qwen 3.7–Max has strong logical reasoning ability, it is not the best choice for high-volume low-complexity tasks where it will take a significant amount of time to Reason internally before proceeding to actual execution. There are some multimodal visual or auditory tasks, especially those being performed in a complex physical environment that will rely on external processing modules via Multi-Agent Pipelines having handoffs.

Future Architectural Enhancements

Could the creators of the model implement dynamic neuro-symbolic scaffolding in the core sparse routing architecture of the algorithm? This would be the direction that can be pursued by the internal research teams responsible for further development of the proprietary solution, moving from fixed parameters to online learning processes. This strategy enables the system to continuously update expert models in real-time without the problem of catastrophic forgetting. In turn, it would enable to drastically improve the performance of baseline inference processes by eliminating heavy offline training cycles.

Moreover, can the architects of the company’s proprietary infrastructure integrate memory checkpoints and standard agent-to-agent communication protocols into the attention mechanism? Instead of relying on external open-source tools that implement prompt-based scaffolding strategies to orchestrate the process, these protocols could be embedded into the cloud execution engine itself, enabling to get rid of the existing latency entirely. Thus, the system could be turned into an organic orchestral solution capable of cross-platform collaboration.

Conclusion

While prioritizing long-term execution stability and format-agnostic interactions with tools over traditional benchmarks, the approach makes a move towards reliable, multi-day digital workers. In today’s production systems, the key aspect changes from managing vulnerable prompt structures to coordinating self-sufficient pipelines that can solve any problem independently.

Source
Blog: https://qwen.ai/blog?id=qwen3.7

Disclaimer - This article is intended purely for informational purposes. It is not sponsored or endorsed by any company or organization, nor does it serve as an advertisement or promotion for any product or service. All information presented is based on publicly available resources and is subject to change. Readers are encouraged to conduct their own research and due diligence.

Cline : Open Source Agentic Ecosystem Across SDK IDE CLI

Introduction

Today’s software engineering necessitates the ability to reliably execute code; hence, revealing the inadequacies of conventional interactive environments that integrate execution logic within the interface surface itself. Ensuring continuous operation of extended code generation processes through application crashes or UI reloads necessitates flexible software design where logic is standardized independently of any particular surface wrapper. Achieving this consistency heavily relies on keeping non-persistent core execution loops along with portable, decoupled life cycle management systems. Moreover, ensuring scalability of sophisticated code modification operations requires intrinsic agent delegation among peers, along with defined programmatic execution contexts.

The introduction of Cline SDK comes at just the right time because of precisely those needs. The SDK separates the core logic of the tool from the rest of the components, allowing for the execution environment to become embeddable in a wide array of interfaces. Integrating the code assistant as an extension of the multi-surface IDE, a CLI tool within your local terminal or a cloud-based CI environment allows one to build up a service-oriented coding environment.

What is Cline?

Cline is a full-fledged agentic ecosystem for engineering, developed by Cline Bot Inc. It is capable of operating as either a programmatic software development kit (@cline/sdk), an integrated development environment (IDE) extension, or as an interactive command-line interface (CLI). Essentially, it acts as an extensible software companion, transforming high-level functional specifications into low-level codebase modifications by means of natural language processing along with secure system tool invocation protocols, and operates as a utility engine which safely complements human software engineering efforts.

Key Features of Cline

An analysis of Cline's technical features suggests that this software was developed with high controllability and safety features in mind. Key architectural capabilities of Cline include:

• Human-in-the-Loop (HITL) Gatekeeping. In order to avoid any destructive impacts of an automatic change, Cline operates using strict security measures when it comes to alterations in the files and command lines, pausing for human confirmation each time such action is needed.
• Real-time environmental analysis: Unlike other systems, Cline continuously analyzes the project workspace by conducting in-depth Abstract Syntax Tree (AST) parsing, regex, and automatic linter/compiler monitoring. Thus, if a code modification leads to broken syntax, types or missing import, Cline finds it and corrects before the task completion.
• Dual cognitive modalities: In order to minimize a token cost and maximize efficiency, the system separates actions into two mental modes. Plan mode is responsible for architecture assessment, structural dependencies' review and asking clarification questions without interfering in the code at all. On the contrary, act mode deals with code execution only.
• Agnostic Model Infrastructure: The infrastructure incorporates an abstraction layer that separates the core large language model from the toolset. This enables switching across more than 200 models including Anthropic, OpenAI, Google Gemini, AWS Bedrock, Azure, and GCP Vertex as well as open-weight execution locally using Ollama or LM Studio.
• Integration of Model Context Protocol (MCP): Cline is different from other toolsets due to the inclusion of MCP servers in the infrastructure. It enables dynamic enhancement of the agent's skills by connecting to secure databases, remote cloud environments or any third-party utility APIs using the open standard protocol.

Use Cases of Cline

• The Secure Air-Gapped Software Factory

In case the organization has strict constraints dictated by certain regulations (defense, financial services infrastructure, health care) the use of code generation tools based on the cloud brings severe compliance risks as well as IP threats. Due to the nature of Cline that is vendor-neutral when it comes to backend execution logic the team can set up their own air-gapped software factory. Using Ollama and LM Studio it will be possible to bind the SDK with local hardware with locally deployed open-weight architectures allowing deep refactoring, patches application, and migrations without sending even a single byte of proprietary code anywhere beyond your network perimeter.

• Multi-Model Agentic Performance Benchmarking

The choice of the best-performing large language model depends on the trade-off between the precision of code generated and the cost and time needed for inference. It's possible to create meta-agents using @cline/llms module to benchmark different providers based on a precise coding task like migrating a legacy service from CommonJS to ECMAScript modules.

• Parallel Agile Task Management with Digital Workforces

The traditional workflow process of AI restricts developers into sequential interactions that form a cognitive bottleneck. By adopting the visual orchestration layer of Cline's Kanban task board (npx kanban), the product managers and technical leads can scale a parallel digital workforce. Every card on the task board is either a feature request or a bug report. Underneath the visual cards, the SDK launches a specialized agent for each task, which runs on its unique worktree and commits separately. One engineer is able to coordinate dozens of parallel agents modifying different parts of the codebase independently.

• Recovery Through Edge Messaging Channels

In cases where there is a system failure that occurs out of regular business hours, the time taken for recovery will be dependent on the time taken by an engineer to physically arrive at his/her computer to address the problem. Cline runtime has channel connectors which allow the access to agents via secure messaging platforms such as Slack, Discord, Telegram, or WhatsApp through cline connect configuration wizard. In case of an incident from a production monitoring alert, an on-call engineer can request a headless Cline agent right from his/her phone messaging application. The agent makes use of the runtime access to diagnose the server logs and generate a clean code diff which is approved by the engineer and kick starts the CI/CD pipeline process.

How Does Cline Work?

Cline 2.0 comes with a strict decoupling and layering TypeScript stack (as shown in figure below) intended to keep single-responsibility separation within its ecosystem. The design breaks down the core into three layers: application interface at the surface layer, stateful runtime and the stateless agent loop, all components depending solely on the layer below. The foundation layer of the engine is called @cline/llms and it fully abstracts the settings, API configurations and token counting for model-specific catalogs. Programmers can easily plug new artificial intelligence backends into the ecosystem by implementing a generic ApiHandler interface making the core engine model agnostic.

source - https://cline.bot/blog/introducing-cline-sdk-the-upgraded-agent-runtime

The actual advantage of this flow is the separation of execution processes from the stateless loop into the stateful runtime wrapper. Having stateless execution at the lowest level enables this software to be easily scaled into an ephemeral serverless deployment scenario as well as being embedded on a micro-surface without dragging any heavy data baggage. The external stateful runtime would take care of the persistence aspects, user sessions, compilation logs, and even file system changes. Such a two-layer execution flow focuses primarily on systemic safety by producing cryptographic checkpoints for each and every edit performed within the codebase in order to allow easy diff inspection and rollbacks.

Performance Evaluation and Benchmarks

The peer-reviewed Terminal Benchmark suite (tbench.ai) was used to measure the performance of Cline's CLI engine according to architectural innovation and its capacity to solve complex, multi-step software engineering tasks.

source - https://cline.bot/blog/introducing-cline-sdk-the-upgraded-agent-runtime

After reviewing the performance of Cline vs existing implementations of both high-level frontier models Cline's improvements have resulted in a significant increase in the efficiency of Cline vs other systems due to the optimization in managing the context. The results of the evaluation of the Cline CLI on the claude-opus-4.7 architecture resulted in a success rate of 74.2% for pass @ 1 success, as opposed to Anthropic's native Claude Code terminal application success rate of 69.4%. The performance difference indicates Cline's proprietary formatting of inputs so as to format codebase contextual information to the methods of reinforcement learning produced results with fewer errors across longer multiple-step tasks. The platform has shown consistent performance across multiple inference engines compared to other model types. Cline scored 71.9% in comparison to other architecturally distributed models, such as Claude Code (65.4%) and Droid (69.9%), while being run on an architecture that uses the claude-opus-4.6 model set. 

source - https://cline.bot/blog/introducing-cline-sdk-the-upgraded-agent-runtime

On distributed architectures that used vanilla (i.e., open-weight) local models, Cline scored 55.1% using a kimi-k2.6 model; in comparison, all other agent models scored less, including OpenCode (37.1%) and Pi-Code (45.5%). For test round evaluations using gpt-5.3-codex on the Cline platform, the score was a 73.0% pass rate, which was comparable to other system-specific models, including the Codex CLI framework (75.1%).

How to Access and Use Cline?

Cline is entirely open-source and distributed under the Apache 2.0 license. That is, the ecosystem can be used commercially without any restrictions and even locally modified and hosted on-premises. The entire source code and all related resources can be found in the official Cline GitHub repository. The whole ecosystem can be installed via standard package managers. For those who wish to develop a custom agent application, the SDK can be easily installed with npm install @cline/sdk. If an interactive terminal workflow is preferred, the command-line helper can be installed globally using 'npm i -g cline' command.

Limitations 

Although the adoption of the modular 2.0 SDK represents an important improvement in terms of stability, there are some aspects of the Cline ecosystem that are still being developed actively. At the moment, the CLI tool and the visualization feature of the Kanban board have successfully been ported to the new 2.0 SDK structure, although moving the VS Code and JetBrains IDE plugins to this architecture is still under progress. There is also an existing disparity within the ecosystem concerning openness as the plugins for the JetBrains product line are not open source as of the moment.

Future Work

The communication connectors designed for routing agent activity via messaging systems beyond the platform (e.g., Slack, Discord, WhatsApp, Telegram) are still under evaluation as a feature of the platform, such that it may result in connection interruptions/failures when deployed within complex companies that utilize proxy servers or under strict security measures within their respective enterprise networking environments. The development team will continue collecting community input and software bugs to improve these architectural issues when scaling up use on multiple surfaces.

Conclusion

Through this new architecture, technology leaders and software developers will change their perception of automation as it relates to engineering. The new architecture moves coding assistance to developers' IDEs (Integrated Development Environments) from their isolated workspaces, directly integrating them into the broader developer infrastructure, establishing an order of magnitude more scalable framework upon which engineering teams can build in today's environment.

Sources:
Blog: https://cline.bot/blog/introducing-cline-sdk-the-upgraded-agent-runtime
GitHub Repo: https://github.com/cline/cline
Document: https://docs.cline.bot/cline-overview

Disclaimer - This article is intended purely for informational purposes. It is not sponsored or endorsed by any company or organization, nor does it serve as an advertisement or promotion for any product or service. All information presented is based on publicly available resources and is subject to change. Readers are encouraged to conduct their own research and due diligence.

Mistral Medium 3.5: 256K Context Multimodal For Cloud Agents

Introduction

Companies around the world are depending more and more on computerized digital technology to cope with complex software development lifecycle issues and conduct goal-directed digital operations independently. At the same time, the ability to handle visual information inputs that aren't well defined, such as graphs and drawings, as well as extract structured formats from raw data, is crucial for maintaining momentum. Engineering groups used to deal with various highly specialized digital programs to accomplish this objective in the past. In contrast, today's solutions incorporate the capabilities of extremely focused and specialized systems in a unified platform. The unification of structure is what allows for transforming high-level technical studies and background operations into useful tools.

One can see how Mistral Medium 3.5, which serves as a perfect illustration of this transformation, responds to the need for a solution that would be able to address multiple problems at once within a single framework. The latest web update demonstrates its use as a foundation for Mistral s Vibe remote coding agents and Le Chat s Work mode, shifting the paradigm from chat aid to delegated cloud computing.

source - https://mistral.ai/news/vibe-remote-agents-mistral-medium-3-5

What is Mistral Medium 3.5? 

Mistral Medium 3.5 (MM3.5) is an extremely dense 128 billion parameter flagship multimodal AI that functions as a unified backend execution system for long-term enterprise workflows. It reduces multiple distinct, specialized domain-specific models – Magistral model designed for deep reasoning, Devstral designed for agentic coding, and Mistral Medium itself for instruction following tasks into one single model capable of handling text and image inputs. Announced towards the end of April 2026, it was designed to be able to function as either an intelligent lightweight assistant or an asynchronous cloud agent for deep thinking tasks, with support for tool calling.

Key Features of Mistral Medium 3.5

• Unified Modality and Extreme Context Ingestion:  MM3.5 can accept multimodal input types, including not only text but also images of arbitrary sizes. The output will be generated as text, too. To process extensive amounts of information, it has an enormous context size of 262,144 tokens (256k). Therefore, the model can examine large repositories of software codes, thorough API documentation, or numerous pages of legal and policy documents all at once, preserving the main story.
• Dynamic, Controllable Reasoning Effort:  An important feature of the model is a unique dynamic reasoning_effort option included in the payload. Users can select either  none  or  high  levels for this parameter. If  none  is selected, then MM3.5 can operate as a fast, small conversational agent. When  high  is selected, the model will use test-time computing resources and work as a deep thinker, ready to solve complicated problems step-by-step.
• Asynchronous Agentic Persistence:  Standard chat applications require the user's browser or terminal to be open throughout the entire conversation. Contrarily, agents based on MM3.5 in Le Chat's  Work  mode or the Vibe CLI can operate independently and continuously until the completion of their task.
• Built-In Enterprise Connectors On by Default: The model frees up users from the tiresome task of manual context collecting. In the Work mode, connections to necessary productivity software such as Gmail, Google Drive, Notion, Slack, and Jira are set up automatically. The agent uses its capabilities to retrieve rich context from these systems to make correct decisions.
• Isolation, Sandboxing, and Scalable Simultaneous Operations: Securely developed, Mistral Medium 3.5 supports simultaneous remote code editing sessions. Each one takes place in an isolated sandbox, allowing the user to freely edit multiple files, refactor modules, and install software without risking to interfere with other agents or cause any harm to his/her hardware.
• Multilingual Proficiency: In order to satisfy global enterprises' needs, the model can work efficiently with dozens of languages. It exhibits excellent fluency and nativeness while using English, French, Spanish, German, Chinese, Japanese, and Arabic, etc.
• Autonomous Transparency: As opposed to the focus on efficiency and speed of the majority of models out there, Mistral Medium 3.5 prioritizes transparency by showing its user the full picture of what is going on inside the system. It discloses every tool call and explains the decision-making process.

Use Cases for Mistral Medium 3.5

• Session Teleportation for Bypassing Hardware Limitations: Gone are the days when hours spent refactoring would tie up local machines. The ability to teleport the session with many tools employed to the cloud-based agent allows computation offloading with no loss of existing context and access rights. This way, the focus moves from tedious source code tweaking to the Pull Request assessment, saving half of the time.
  
• Saving on Maintenance Expenses: Scaling requires an ecosystem that sustains itself. The model’s ability to generate and merge 90% of its own platform PRs allows its deployment into practical incident monitoring platforms. It automatically deals with broken CI pipelines and applies patches in the background. As such, it covers the expenses connected to maintenance, leaving people free to work only on designing the architecture.
  
• Deploying Flagship AI in Heavily Regulated Industries: Enterprises with highly sensitive data do not have the option of relying on third-party API calls, but running unpredictable Mixture-of-Expert models internally requires substantial investment in hardware infrastructure. Since this is a highly compact and predictable 128B model, world-class AI solutions can run behind firewalls using only four ordinary GPUs. The end result will be complete data sovereignty and total predictability in capacity planning and hardware costs.
  
• Meeting Global Compliance Standards in Non-English-speaking Countries: Autonomous agents require assurance of certainty that internal logic corresponds to actions in order to create an audit trail. While most approaches are characterized by language mixing, where agents use English first before translating, this particular approach actively discourages this kind of behavior through learning processes. This assures complete compliance and auditability in environments using Arabic, Russian, or Chinese languages by ensuring that internal logic and actions are conducted in their native languages.
  
• Substantial Increase in System Performance in CI/CD Pipelines: Automating the management of a large number of tasks or conducting immediate triaging necessitates fast processing speeds to prevent potential bottlenecks. While most deep reasoning models require long periods to process tasks, combining this model with its EAGLE variant will increase its processing speed two-fold. It will provide instant services capable of handling complicated requests on the spot without compromising intelligence levels for success.

What Is the Process Behind Mistral Medium 3.5?

The Mistral Medium 3.5 leverages a 128-B-parameter dense Transformer architecture. The intentional move from a sparse Mixture-of-Experts (MoE) approach guarantees that the model has an uncontaminated vocabulary embedding and deterministic execution backend for long-horizon agentic operations. For effective processing of visuals, the model abandons its inherited universal encoders and builds a custom one from scratch. This custom module is specially designed to cater to images of different dimensions and aspect ratios, increasing the accuracy of Mistral's visual reasoning in comprehending unstructured data like unconventional documents, user interface snapshots, and complicated architectural drawings.

The working mechanism involves developing the model through a Control Plane locally (Vibe CLI) and an Execution Plane cloud-side (agents remotely controlled through Mistral Studio Workflows). In terms of efficiency, the base model works best when coupled with the EAGLE speculator version of the model. When generating content, the drafting model repeatedly inputs predicted tokens into the 128B model, which evaluates the inputted batches using its self-attention layers in one go to either approve or deny the prediction. With the asynchronous reinforcement learning pipeline using fastText classification, the system improves its efficiency without affecting the user's session parameters.

Performance Evaluation with Other Models

The Mistral Medium 3.5 has exhibited absolute supremacy in the automated software engineering industry in the extremely rigorous industry evaluation charts. In one of its key tests, SWE-Bench Verified, it earned 77.6%. The significance of such a score is that it reflects a large improvement from its code generator variant, Devstral 2 (72.2%), and outperforms the state-of-the-art models, Anthropic s Claude Sonnet 4.5 (77.2%) and Qwen3.5 397B A17B (76.4%). This is because, in this test, the capabilities of the model are evaluated on whether it can solve problems in the GitHub ecosystem autonomously.

source - https://mistral.ai/news/vibe-remote-agents-mistral-medium-3-5

Furthermore, when tested on multi-step orchestration performance, the model demonstrated yet another success by achieving 91.4 in the tau3-Telecom agentic test. This particular test evaluates the capabilities of a model in calling tools reliably and executing long-horizon workflows. With such a high score, the Mistral Medium 3.5 proves itself to rarely hallucinate inputs to its tools. Hence, it becomes the accurate model for asynchronous human-less cloud agents.

How to Access Mistral Medium 3.5?

The Mistral Medium 3.5 is instantly downloadable from the Hugging Face page as open weights. It comes as the native implementation of the default execution engine behind the  Work mode  function of the Le Chat application and Vibe CLI. In enterprise environments, the Mistral Medium 3.5 is accessible through the Mistral AI Studio API and provided as an NVIDIA NIM package. To run the model in-house, developers can refer to the detailed guidelines in the GitHub repository of high-performance inference systems like vLLM, SGLang, and llama.cpp. The model is released under a Modified MIT License, which is still very liberal in terms of usage rights and allows its free usage in both business and personal capacities, except for corporations that earn vast sums globally.

Limitations 

While being groundbreaking in design terms, there are several real-life limitations this AI operates under. Firstly, since it works under a modified MIT license, which does not allow completely unrestricted use, big corporate clients have to negotiate their own custom commercial license agreements. Secondly, while in terms of design, the AI is created specifically for long runs, which it executes through a giant 256k context window, empirical research shows that for contexts longer than 40,000 tokens, reasoning accuracy may decrease at some point.

Future Work

Looking ahead into the future, the team at Mistral AI has made it clear that they have hired people in order to take these agentic systems even further, implying that in the future versions, emphasis would be placed on further developing autonomous decision-making capabilities.

Conclusion

The real value of the release of Mistral Medium 3.5 lies not only in the sheer density of its parameters, but in the understanding that with a seamlessly integrated cloud-to-local system, backed by state teleportation and speculative decoding via EAGLE, time can literally be cut down in half. Technical decision-makers who wish to create their own autonomous triage systems should consider using a predictable-compute system that created its own infrastructure as their safest possible bet.

Sources:
Blog: https://mistral.ai/news/vibe-remote-agents-mistral-medium-3-5
Model Weight: https://huggingface.co/mistralai/Mistral-Medium-3.5-128B
Model Card: https://docs.mistral.ai/models/model-cards/mistral-medium-3-5-26-04
Model Guide: https://docs.mistral.ai/models/model-selection-guide?models=mistral-medium-3-5-26-04
Eagle Model: https://huggingface.co/mistralai/Mistral-Medium-3.5-128B-EAGLE

Disclaimer - This article is intended purely for informational purposes. It is not sponsored or endorsed by any company or organization, nor does it serve as an advertisement or promotion for any product or service. All information presented is based on publicly available resources and is subject to change. Readers are encouraged to conduct their own research and due diligence.

DeepSeek-V4: Low-Cost Logic via Hybrid Attention Architectures

Introduction

There is an evident inclination toward novel innovations that incorporate unique structural modifications in modern sparse neural models. Specifically, this approach incorporates multi-attention principles capable of addressing large volumes of informational throughput. Simultaneously, an increased need for native implementation of autonomous algorithms in sophisticated STEM applications and long-term operations is noted. By incorporating specialized data, obtained by experts within highly focused domains, into one engine, this innovation becomes exceptionally affordable.

In this light, a new AI architecture emerges as a highly cost-optimized framework. It is specially created in order to integrate 1-million-token contexts into its design automatically. In essence, it makes the application of large context lengths free of charge for companies constructing efficient scalable systems. It changes the very economics behind extended agent operations. Further analysis will shed light on the mechanics of operation, various versions of deployment, advanced functionalities and benchmarks of the innovation. This particular innovation is known as 'DeepSeek-V4'.

What is DeepSeek-V4?

DeepSeek-V4 is a highly optimized Mixture-of-Experts (MoE) large language model designed to achieve ultra-high computational efficiency for million-token context processing. By rethinking how attention mechanisms and residual connections operate, it establishes a new baseline where maintaining massive amounts of conversational and reasoning history is handled with drastically reduced compute and memory costs, enabling persistent, long-horizon digital operations without degrading performance.

Model Variants

• DeepSeek-V4-Pro (1.6 trillion Total Parameters / 49 billion Active Parameters): The Pro design sets new benchmarks for open-weights models. The design is optimized to perform the most challenging logic, mathematics, and programming tasks. With its development being slightly behind proprietary frontier models by a few months, DeepSeek offers enterprise-level reasoning abilities for complex, multi-stage problems that need utmost precision.
• DeepSeek-V4-Flash (284 billion Total Parameters / 13 billion Active Parameters): Designed for unparalleled speed and maximum efficiency, the Flash model boasts high parameter efficiency. While delivering better performance than the earlier V3.2-Base model with far fewer requirements, DeepSeek achieves nearly identical reasoning accuracy as the Pro model when provided with more computing power.

Modes of Reasoning Effort

• The Non-Think Mode is optimized for use with routine tasks and/or low-risk decisions, providing fast, intuitive output.
• The Think-High Mode uses the 128K context window to enable users of the program to perform conscious logical reasoning and deep planning or multiple steps of tool use.
• The Think-Max Mode is a boundary-expanding context window setting where 384K tokens are required. The Think-Max Mode has a specialized system prompt to utilize a maximum level of recursion, decomposing complex numerical and logical problems into the most minute of detail for the highest level of mathematical and logical research possible.

Key Features of DeepSeek-V4

The design brings multiple structural improvements that positively affect the cost of deployment and inference.

• Extreme Efficiency in Handling Long Contexts: Working with a large amount of context (such as 1M tokens) typically results in significant  context decay  issues. In contrast, DeepSeek-V4-Pro consumes 27% FLOPs and 10% KV cache compared to DeepSeek-V3.2, while DeepSeek-V4-Pro Flash consumes an astonishingly low 10% of FLOPs and 7% KV cache.
• Persistent Interleaved Reasoning: The earlier designs tended to drop any internal reasoning traces once a new input was received from users or outputs from tools. V4 maintains the entire set of traces during the whole conversation intrinsically. Hence, all long-horizon agentic actions have a perfect continuity of planning processes regardless of their number.
• Short Instruction Handling Using Auxiliary Tokens: V4 has introduced several special tokens such as "<|action|>, <|query|>, <|title|>" and "<|authority|>". Adding them to any input would allow the model to use KV cache to execute auxiliary tasks such as intent recognition or search generation without prefilling.
• Agentic Search and Tool Call Using XML Format: During the thought process, V4 uses Agentic Search instead of conventional RAG, which enables the model to repeatedly call the tool to handle difficult questions without increasing costs significantly. Moreover, it employs a new XML format that uses the |DSML| token to minimize escaping problems when executing tools.

Use Cases of DeepSeek-V4

The following examples make use of the distinct advantages of the V4 architecture, which are entirely novel compared to other competing architectures in the market.

• Deterministic Task Resumption in Agentic, Cluster-wide Workflows
  Even in large-scale computer clusters, failures of hardware components are inevitable. By using token-level Write-Ahead Log (WAL) that stores the state of generation and KV caches, V4 allows a multi-hour long mission-critical process to start again where it was left off after the interruption. Such an approach saves millions of computational cycles wasted and minimizes mathematical bias that is inherent to restarting generation from scratch.
• Persistent Thought-based Refactoring of Legacy Codebase across Multiple Sessions

Consider a hypothetical scenario where a large-scale migration of the multi-million lines of code in a legacy code base needs to be done into the latest microservices architecture paradigm. With deep seeking V4 having a capability of Interleaved Thinking Persistence inbuilt, there would be no way that previous reasoning traces can be discarded across thousands of calls to tools. With architectural optimizations that allow execution within a small memory footprint, i.e., 10% of normal KV cache usage, the high fidelity persistence over 1M-token spans would become feasible without any risks of triggering Out-Of-Memory exceptions.

• Prototype Development of Custom Attention Kernels using SMT Verification

In laboratories interested in developing custom sparse attention layers for specialized industries, V4 offers tremendous advantages in its environment due to TileLang being a dedicated language that includes an SMT-solver (Z3). Thus, quick prototyping of attention layers with integer formal analysis becomes possible along with automatic detection of memory issues making kernels memory-safe for trillions of parameters.

• Acquiring Formal Logic for Advanced Mathematics
  

Automated creation of proofs for advanced mathematics entails reasoning ability that stretches the bounds of computational capability. By putting V4 into  Think Max  mode, which demands a context window size that exceeds 384K, the program is compelled to reason on the edge through recursive breakdown of the problems. This makes the software perfect for validating mathematical proofs, both informal and formal.

How Does DeepSeek-V4 Work?

In terms of the inner workings of V4, it is far beyond conventional architectures in that it adopts a Hybrid Attention Mechanism. It combines Compressed Sparse Attention (CSA), wherein compression is carried out at a ratio of m while sparse attention is used on top k entries with Heavily Compressed Attention (HCA), in which the degree of compression is more extreme to group entries with dense attention. To prevent signal decay due to the great depth in terms of number of parameters, traditional residual connections are substituted with Manifold-Constrained Hyper-Connections (mHC).

source - https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/resolve/main/DeepSeek_V4.pdf

Moreover, stability and optimization processes have undergone immense changes. While the hidden layers rely on Muon optimizer to achieve faster convergence, loss spikes are prevented by the means of Anticipatory Routing (which involves calculation of routing indices based on historical parameters) and SwiGLU Clamping (linear components are bound to a value range of [-10, 10]). In terms of hardware improvements, Expert Parallelism (EP) Mega Kernel ensures full overlap of computation and communication processes for the sake of 1.96X latency reduction in rollouts. Lossless dequantization to FP8 is performed on MoE expert weights during Quantization Aware Training (QAT) in the form of conversion from FP4 representation. Finally, On-Policy Distillation (OPD) process is applied, which comprises two stages involving training of domain experts prior to multi-teacher logit-level distillation.

Performance Evaluation with Other Models

From Table below in the performance metrics for the model, DeepSeek-V4 sets another historical record for formal reasoning and mathematics, obtaining the perfect mark of 120 out of 120 in the Putnam-2025 competition. 

source - https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/resolve/main/DeepSeek_V4.pdf

It did so using the combination of informal reasoning and strict formal verification. The perfect mark obtained means a lot since DeepSeek-V4 is able to use its mastery of complex multi-level decomposition of problems without getting itself involved in logical hallucination.

source - https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/resolve/main/DeepSeek_V4.pdf

In addition, from Table above showing performance results in coding competitions, DeepSeek-V4-Pro-Max has the same level of coding ability as GPT-5.4. It has marked a historical moment when an open-weights model is able to compete at par with a closed-source frontier model in this specific domain. In the global Codeforces ranking system, DeepSeek-V4 holds the 23rd position among all humans.

How to Access and Use DeepSeek-V4?

DeepSeek-V4 is freely accessible and usable at chat.deepseek.com in both modes of Expert and Instant with direct integration capabilities provided by the DeepSeek API that is compatible with OpenAI and Anthropic formats. Model weight files in both flash and pro versions are freely accessible via the Hugging Face website, thereby providing deployment options locally or privately on your server. It should be noted that official support for deepseek-chat and deepseek-reasoner will cease from July 24, 2026, henceforth routing traffic to DeepSeek-V4 Flash.

Limitations 

Firstly, the V4 architecture is known at the moment for its complexity because of the application of lots of newly proven tricks related to structural architecture, which should be improved further and made more concise in the future. Furthermore, the Flash model is not equal in terms of the number of parameters in comparison with the Pro variant, thus, having less knowledge about the world than Pro; besides, there is still the necessity for the model to improve its formatting aesthetics to manage specific tasks, such as slide creation and summarizing extreme text.

Future Frontiers: Adaptive Kernels & Memory Meshes
Onwards, what potential may be unlocked with the introduction of Hardware-Aware Self-Compiling Kernels on top of the current efficiency offered by the sparse architecture? With the help of the already-existing formal verification methodology, the system may dynamically compile new attention kernels to utilize certain memory hierarchy structures available in future hardware like Blackwell or even customized edge accelerators. This self-optimization may unlock an almost seamless transition between ultra-precise reasoning and under one millisecond of response time for horizons up to one million tokens.

Additionally, there exists huge potential of expanding session-based persistence into a full-fledged Distributed Agentic Memory Mesh. As opposed to isolated traces of reasoning, will it be possible to develop a federated layer where multiple agents utilize the same live KV cache distributed across a set of nodes in a cluster? This way, it will be possible to create a true collaboration platform, a Thinking Cloud that performs massive overhauls orchestrated by a fleet of agents while sustaining the correct trajectory of reasoning without any extra prefilled information.

Conclusion

By cutting the cost of processing 1-million-token window dramatically and providing the opportunity to use true token-level fault tolerance through Write-Ahead Log, it connects experimental AI to rock-solid enterprise infrastructure. Considering the direction of development of digital ecosystems as persistent thinkers, V4 provides an adequate foundation.

Sources:
Blog: https://api-docs.deepseek.com/news/news260424
API document: https://api-docs.deepseek.com/news/news260424
Tech Document: https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/resolve/main/DeepSeek_V4.pdf
Model Variants: https://huggingface.co/collections/deepseek-ai/deepseek-v4
Model weight Flash: https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash
Model weight Pro: https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro

Disclaimer - This article is intended purely for informational purposes. It is not sponsored or endorsed by any company or organization, nor does it serve as an advertisement or promotion for any product or service. All information presented is based on publicly available resources and is subject to change. Readers are encouraged to conduct their own research and due diligence.