Researchers at Alibaba’s Tongyi Lab have developed a new framework for self-evolving agents that create their own training data by exploring their application environments. The framework, <a href="https://github.com/modelscope/AgentEvolver">AgentEvolver</a>, uses the knowledge and reasoning capabilities of large language models for autonomous learning, addressing the high costs and manual effort typically required to gather task-specific datasets.Experiments show that compared to traditional reinforcement learning–based frameworks, AgentEvolver is more efficient at exploring its environment, makes better use of data, and adapts faster to application environments. For the enterprise, this is significant because it lowers the barrier to training agents for bespoke applications, making powerful, custom AI assistants more accessible to a wider range of organizations.<h2>The high cost of training AI agents</h2><a href="https://venturebeat.com/ai/open-source-deepseek-r1-uses-pure-reinforcement-learning-to-match-openai-o1-at-95-less-cost/">Reinforcement learning</a> has become a major paradigm for training LLMs to act as agents that can interact with digital environments and learn from feedback. However, developing agents with RL faces fundamental challenges. First, gathering the necessary training datasets is often prohibitively expensive, requiring significant manual labor to create examples of tasks, especially in novel or proprietary software environments where there are no available off-the-shelf datasets.Second, the RL techniques commonly used for LLMs require the model to run through a massive number of trial-and-error attempts to learn effectively. This process is computationally costly and inefficient. As a result, training capable LLM agents through RL remains laborious and expensive, limiting their deployment in custom enterprise settings.<h2>How AgentEvolver works</h2>The main idea behind AgentEvolver is to give models greater autonomy in their own learning process. The researchers describe it as a “self-evolving agent system” designed to “achieve autonomous and efficient capability evolution through environmental interaction.” It uses the reasoning power of an LLM to create a self-training loop, allowing the agent to continuously improve by directly interacting with its target environment without needing predefined tasks or reward functions.“We envision an agent system where the LLM actively guides exploration, task generation, and performance refinement,” the researchers wrote in <a href="https://arxiv.org/abs/2511.10395">their paper</a>.The self-evolution process is driven by three core mechanisms that work together.The first is self-questioning, where the agent explores its environment to discover the boundaries of its functions and identify useful states. It’s like a new user clicking around an application to see what’s possible. Based on this exploration, the agent generates its own diverse set of tasks that align with a user’s general preferences. This reduces the need for handcrafted datasets and allows the agent and its tasks to co-evolve, progressively enabling it to handle more complex challenges. According to Yunpeng Zhai, researcher at Alibaba and co-author of the paper, who spoke to VentureBeat, the self-questioning mechanism effectively turns the model from a “data consumer into a data producer,” dramatically reducing the time and cost required to deploy an agent in a proprietary environment.The second mechanism is self-navigating, which improves exploration efficiency by reusing and generalizing from past experiences. AgentEvolver extracts insights from both successful and unsuccessful attempts and uses them to guide future actions. For example, if an agent tries to use an API function that doesn&#x27;t exist in an application, it registers this as an experience and learns to verify the existence of functions before attempting to use them in the future.The third mechanism, self-attributing, enhances learning efficiency by providing more detailed feedback. Instead of just a final success or failure signal (a common practice in RL that can result in sparse rewards), this mechanism uses an LLM to assess the contribution of each individual action in a multi-step task. It retrospectively determines whether each step contributed positively or negatively to the final outcome, giving the agent fine-grained feedback that accelerates learning. This is crucial for regulated industries where how an agent solves a problem is as important as the result. “Instead of rewarding a student only for the final answer, we also evaluate the clarity and correctness of each step in their reasoning,” Zhai explained. This improves transparency and encourages the agent to adopt more robust and auditable problem-solving patterns.“By shifting the training initiative from human-engineered pipelines to LLM-guided self-improvement, AgentEvolver establishes a new paradigm that paves the way toward scalable, cost-effective, and continually improving intelligent systems,” the researchers state.The team has also developed a practical, end-to-end training framework that integrates these three mechanisms. A key part of this foundation is the Context Manager, a component that controls the agent&#x27;s memory and interaction history. While today&#x27;s benchmarks test a limited number of tools, real enterprise environments can involve thousands of APIs. Zhai acknowledges this is a core challenge for the field, but notes that AgentEvolver was designed to be extended. “Retrieval over extremely large action spaces will always introduce computational challenges, but AgentEvolver’s architecture provides a clear path toward scalable tool reasoning in enterprise settings,” he said.<h2>A more efficient path to agent training</h2>To measure the effectiveness of their framework, the researchers tested it on <a href="https://appworld.dev/">AppWorld</a> and <a href="https://gorilla.cs.berkeley.edu/blogs/13_bfcl_v3_multi_turn.html">BFCL v3</a>, two benchmarks that require agents to perform long, multi-step tasks using external tools. They used models from Alibaba’s <a href="https://venturebeat.com/ai/qwen-swings-for-a-double-with-2-5-omni-3b-model-that-runs-on-consumer-pcs-laptops">Qwen2.5 family</a> (7B and 14B parameters) and compared their performance against a baseline model trained with GRPO, a popular RL technique used to develop reasoning models like <a href="https://venturebeat.com/ai/deepseek-r1-0528-arrives-in-powerful-open-source-challenge-to-openai-o3-and-google-gemini-2-5-pro">DeepSeek-R1</a>.The results showed that integrating all three mechanisms in AgentEvolver led to substantial performance gains. For the 7B model, the average score improved by 29.4%, and for the 14B model, it increased by 27.8% over the baseline. The framework consistently enhanced the models&#x27; reasoning and task-execution capabilities across both benchmarks. The most significant improvement came from the self-questioning module, which autonomously generates diverse training tasks and directly addresses the data scarcity problem.The experiments also demonstrated that AgentEvolver can efficiently synthesize a large volume of high-quality training data. The tasks generated by the self-questioning module proved diverse enough to achieve good training efficiency even with a small amount of data. For enterprises, this provides a path to creating agents for bespoke applications and internal workflows while minimizing the need for manual data annotation. By providing high-level goals and letting the agent generate its own training experiences, organizations can develop custom AI assistants more simply and cost-effectively.“This combination of algorithmic design and engineering pragmatics positions AgentEvolver as both a research vehicle and a reusable foundation for building adaptive, tool-augmented agents,” the researchers conclude.Looking ahead, the ultimate goal is much bigger. “A truly ‘singular model’ that can drop into any software environment and master it overnight is certainly the holy grail of agentic AI,” Zhai said. “We see AgentEvolver as a necessary step in that direction.” While that future still requires breakthroughs in model reasoning and infrastructure, self-evolving approaches are paving the way.

قدم الباحثون في مختبر تونغيي التابع لشركة علي بابا إطار عمل جديد يسمى AgentEvolver، الذي يمكّن الوكلاء من تطوير أنفسهم من خلال إنشاء بيانات تدريب خاصة بهم من خلال استكشاف بيئاتهم. تشير التجارب إلى أن هذا الابتكار يعزز أداء النموذج في استخدام الأدوات بنسبة تقارب 30% مقارنة بالطرق التقليدية.

Los investigadores del Tongyi Lab de Alibaba han presentado AgentEvolver, un marco que permite a los agentes autoevolutivos generar de manera autónoma sus propios datos de entrenamiento al explorar sus entornos. Esta innovación mejora supuestamente el rendimiento del modelo en el uso de herramientas en aproximadamente un 30 % en comparación con los métodos tradicionales.

Les chercheurs du Tongyi Lab d'Alibaba ont introduit AgentEvolver, un cadre qui permet aux agents auto-évolutifs de générer de manière autonome leurs propres données d'entraînement en explorant leurs environnements. Cette innovation améliore apparemment la performance des modèles d'utilisation d'outils d'environ 30 % par rapport aux méthodes traditionnelles.

Researchers at Alibaba’s Tongyi Lab have introduced AgentEvolver, a framework that enables self-evolving agents to autonomously generate their own training data by exploring their environments. This innovation reportedly enhances model performance in tool use by approximately 30% compared to traditional methods.

Alibaba's AgentEvolver lifts model performance in tool use by ~30% using synthetic, auto-generated tasks

<a href="https://mistral.ai/">Mistral AI</a>, Europe&#x27;s most prominent artificial intelligence startup, is releasing its most ambitious product suite to date: a family of 10 open-source models designed to run everywhere from smartphones and autonomous drones to enterprise cloud systems, marking a major escalation in the company&#x27;s challenge to both U.S. tech giants and surging Chinese competitors.The <a href="https://mistral.ai/news/mistral-3">Mistral 3 family</a>, launching today, includes a new flagship model called <a href="https://mistral.ai/news/mistral-3">Mistral Large 3</a> and a suite of smaller &quot;<a href="https://mistral.ai/models">Ministral 3</a>&quot; models optimized for edge computing applications. All models will be released under the permissive Apache 2.0 license, allowing unrestricted commercial use — a sharp contrast to the closed systems offered by <a href="https://openai.com/">OpenAI</a>, <a href="https://www.google.com/">Google</a>, and <a href="https://www.anthropic.com/">Anthropic</a>.The release is a pointed bet by Mistral that the future of artificial intelligence lies not in building ever-larger proprietary systems, but in offering businesses maximum flexibility to customize and deploy AI tailored to their specific needs, often using smaller models that can run without cloud connectivity.&quot;The gap between closed and open source is getting smaller, because more and more people are contributing to open source, which is great,&quot; Guillaume Lample, Mistral&#x27;s chief scientist and co-founder, said in an exclusive interview with VentureBeat. &quot;We are catching up fast.&quot;<h2>Why Mistral is choosing flexibility over frontier performance in the AI race</h2>The strategic calculus behind <a href="https://mistral.ai/news/mistral-3">Mistral 3</a> diverges sharply from recent model releases by industry leaders. While <a href="https://openai.com/">OpenAI</a>, <a href="https://www.google.com/">Google</a>, and <a href="https://www.anthropic.com/">Anthropic</a> have focused recent launches on increasingly capable &quot;agentic&quot; systems — AI that can autonomously execute complex multi-step tasks — Mistral is prioritizing breadth, efficiency, and what Lample calls &quot;distributed intelligence.&quot;Mistral <a href="https://mistral.ai/models">Large 3</a>, the flagship model, employs a <a href="https://huggingface.co/blog/moe">Mixture of Experts</a> architecture with 41 billion active parameters drawn from a total pool of 675 billion parameters. The model can process both text and images, handles context windows up to 256,000 tokens, and was trained with particular emphasis on non-English languages — a rarity among frontier AI systems.&quot;Most AI labs focus on their native language, but Mistral Large 3 was trained on a wide variety of languages, making advanced AI useful for billions who speak different native languages,&quot; the company said in a statement reviewed ahead of the announcement.But the more significant departure lies in the <a href="https://mistral.ai/models">Ministral 3</a> lineup: nine compact models across three sizes (14 billion, 8 billion, and 3 billion parameters) and three variants tailored for different use cases. Each variant serves a distinct purpose: base models for extensive customization, instruction-tuned models for general chat and task completion, and reasoning-optimized models for complex logic requiring step-by-step deliberation.The smallest Ministral 3 models can run on devices with as little as 4 gigabytes of video memory using 4-bit quantization — making frontier AI capabilities accessible on standard laptops, smartphones, and embedded systems without requiring expensive cloud infrastructure or even internet connectivity. This approach reflects Mistral&#x27;s belief that AI&#x27;s next evolution will be defined not by sheer scale, but by ubiquity: models small enough to run on drones, in vehicles, in robots, and on consumer devices.<h2>How fine-tuned small models beat expensive large models for enterprise customers</h2>Lample&#x27;s comments reveal a business model fundamentally different from that of closed-source competitors. Rather than competing primarily on benchmark performance, Mistral is targeting enterprise customers frustrated by the cost and inflexibility of proprietary systems.&quot;Sometimes customers say, &#x27;Is there a use case where the best closed-source model isn&#x27;t working?&#x27; If that&#x27;s the case, then they&#x27;re essentially stuck,&quot; Lample explained. &quot;There&#x27;s nothing they can do. It&#x27;s the best model available, and it&#x27;s not working out of the box.&quot;This is where Mistral&#x27;s approach diverges. When a generic model fails, the company deploys engineering teams to work directly with customers, analyzing specific problems, creating synthetic training data, and fine-tuning smaller models to outperform larger general-purpose systems on narrow tasks.&quot;In more than 90% of cases, a small model can do the job, especially if it&#x27;s fine-tuned. It doesn&#x27;t have to be a model with hundreds of billions of parameters, just a 14-billion or 24-billion parameter model,&quot; Lample said. &quot;So it&#x27;s not only much cheaper, but also faster, plus you have all the benefits: you don&#x27;t need to worry about privacy, latency, reliability, and so on.&quot;The economic argument is compelling. Multiple enterprise customers have approached Mistral after building prototypes with expensive closed-source models, only to find deployment costs prohibitive at scale, according to Lample.&quot;They come back to us a couple of months later because they realize, &#x27;We built this prototype, but it&#x27;s way too slow and way too expensive,&#x27;&quot; he said.<h2>Where Mistral 3 fits in the increasingly crowded open-source AI market</h2>Mistral&#x27;s release comes amid fierce competition on multiple fronts. OpenAI recently released <a href="https://openai.com/index/gpt-5-1-codex-max/">GPT-5.1</a> with enhanced agentic capabilities. Google launched <a href="https://blog.google/products/gemini/gemini-3/">Gemini 3</a> with improved multimodal understanding. Anthropic released <a href="https://www.anthropic.com/news/claude-opus-4-5">Opus 4.5</a> on the same day as this interview, with similar agent-focused features.But Lample argues those comparisons miss the point. &quot;It&#x27;s a little bit behind. But I think what matters is that we are catching up fast,&quot; he acknowledged regarding performance against closed models. &quot;I think we are maybe playing a strategic long game.&quot;That long game involves a different competitive set: primarily open-source models from Chinese companies like <a href="https://www.deepseek.com/">DeepSeek</a> and Alibaba&#x27;s <a href="https://qwen.ai/home">Qwen</a> series, which have made remarkable strides in recent months.Mistral differentiates itself through multilingual capabilities that extend far beyond English or Chinese, multimodal integration handling both text and images in a unified model, and what the company characterizes as superior customization through easier fine-tuning.&quot;One key difference with the models themselves is that we focused much more on multilinguality,&quot; Lample said. &quot;If you look at all the top models from [Chinese competitors], they&#x27;re all text-only. They have visual models as well, but as separate systems. We wanted to integrate everything into a single model.&quot;The multilingual emphasis aligns with Mistral&#x27;s broader positioning as a European AI champion focused on digital sovereignty — the principle that organizations and nations should maintain control over their AI infrastructure and data.<h2>Building beyond models: Mistral&#x27;s full-stack enterprise AI platform strategy</h2>Mistral 3&#x27;s release builds on an increasingly comprehensive enterprise AI platform that extends well beyond model development. The company has assembled a full-stack offering that differentiates it from pure model providers.Recent product launches include <a href="https://mistral.ai/news/agents-api">Mistral Agents API</a>, which combines language models with built-in connectors for code execution, web search, image generation, and persistent memory across conversations; <a href="https://mistral.ai/news/magistral">Magistral</a>, the company&#x27;s reasoning model designed for domain-specific, transparent, and multilingual reasoning; and <a href="https://mistral.ai/news/mistral-code">Mistral Code</a>, an AI-powered coding assistant bundling models, an in-IDE assistant, and local deployment options with enterprise tooling.The consumer-facing <a href="https://mistral.ai/products/le-chat">Le Chat assistant</a> has been enhanced with Deep Research mode for structured research reports, voice capabilities, and Projects for organizing conversations into context-rich folders. More recently, Le Chat gained a connector directory with 20+ enterprise integrations powered by the Model Context Protocol (MCP), spanning tools like Databricks, Snowflake, GitHub, Atlassian, Asana, and Stripe.In October, Mistral unveiled <a href="https://mistral.ai/products/ai-studio">AI Studio</a>, a production AI platform providing observability, agent runtime, and AI registry capabilities to help enterprises track output changes, monitor usage, run evaluations, and fine-tune models using proprietary data.Mistral now positions itself as a full-stack, global enterprise AI company, offering not just models but an application-building layer through AI Studio, compute infrastructure, and forward-deployed engineers to help businesses realize return on investment.<h2>Why open source AI matters for customization, transparency and sovereignty</h2>Mistral&#x27;s commitment to open-source development under permissive licenses is both an ideological stance and a competitive strategy in an AI landscape increasingly dominated by closed systems.Lample elaborated on the practical benefits: &quot;I think something that people don&#x27;t realize — but our customers know this very well — is how much better any model can actually improve if you fine tune it on the task of interest. There&#x27;s a huge gap between a base model and one that&#x27;s fine-tuned for a specific task, and in many cases, it outperforms the closed-source model.&quot;The approach enables capabilities impossible with closed systems: organizations can fine-tune models on proprietary data that never leaves their infrastructure, customize architectures for specific workflows, and maintain complete transparency into how AI systems make decisions — critical for regulated industries like finance, healthcare, and defense.This positioning has attracted government and public sector partnerships. The company launched &quot;<a href="https://mistral.ai/news/ai-for-citizens">AI for Citizens</a>&quot; in July 2025, an initiative to &quot;help States and public institutions strategically harness AI for their people by transforming public services&quot; and has secured strategic partnerships with France&#x27;s army and job agency, Luxembourg&#x27;s government, and various European public sector organizations.<h2>Mistral&#x27;s transatlantic AI collaboration goes beyond European borders</h2>While Mistral is frequently characterized as Europe&#x27;s answer to OpenAI, the company views itself as a transatlantic collaboration rather than a purely European venture. The company has teams across both continents, with co-founders spending significant time with customers and partners in the United States, and these models are being trained in partnerships with U.S.-based teams and infrastructure providers.This transatlantic positioning may prove strategically important as geopolitical tensions around AI development intensify. The recent ASML investment, a <a href="https://www.nytimes.com/2025/09/09/business/asml-mistral-ai-chips-investment.html">€1.7 billion ($1.5 billion) funding round</a> led by the Dutch semiconductor equipment manufacturer, signals deepening collaboration across the Western semiconductor and AI value chain at a moment when both Europe and the United States are seeking to reduce dependence on Chinese technology.Mistral&#x27;s investor base reflects this dynamic: the Series C round included participation from U.S. firms <a href="https://a16z.com/">Andreessen Horowitz</a>, <a href="https://www.generalcatalyst.com/">General Catalyst</a>, <a href="https://www.lightspeedhq.com/">Lightspeed</a>, and <a href="https://www.indexventures.com/">Index Ventures</a> alongside European investors like France&#x27;s state-backed Bpifrance and global players like DST Global and Nvidia.Founded in May 2023 by former Google DeepMind and Meta researchers, Mistral has raised roughly $1.05 billion (€1 billion) in funding. The company was valued at <a href="https://techcrunch.com/2024/06/11/paris-based-ai-startup-mistral-ai-raises-640-million/">$6 billion</a> in a June 2024 Series B, then <a href="https://mistral.ai/news/mistral-ai-raises-1-7-b-to-accelerate-technological-progress-with-ai">more than doubled its valuation</a> in a September Series C.<h2>Can customization and efficiency beat raw performance in enterprise AI?</h2>The <a href="https://mistral.ai/news/mistral-3">Mistral 3 </a>release crystallizes a fundamental question facing the AI industry: Will enterprises ultimately prioritize the absolute cutting-edge capabilities of proprietary systems, or will they choose open, customizable alternatives that offer greater control, lower costs, and independence from big tech platforms?Mistral&#x27;s answer is unambiguous. The company is betting that as AI moves from prototype to production, the factors that matter most shift dramatically. Raw benchmark scores matter less than total cost of ownership. Slight performance edges matter less than the ability to fine-tune for specific workflows. Cloud-based convenience matters less than data sovereignty and edge deployment.It&#x27;s a wager with significant risks. Despite Lample&#x27;s optimism about closing the performance gap, Mistral&#x27;s models still trail the absolute frontier. The company&#x27;s revenue, while growing, reportedly remains modest relative to its nearly $14 billion valuation. And competition intensifies from both well-funded Chinese rivals making remarkable open-source progress and U.S. tech giants increasingly offering their own smaller, more efficient models.But if Mistral is right — if the future of AI looks less like a handful of cloud-based oracles and more like millions of specialized systems running everywhere from factory floors to smartphones — then the company has positioned itself at the center of that transformation.The release of <a href="https://mistral.ai/news/mistral-3">Mistral 3 </a>is the most comprehensive expression yet of that vision: 10 models, spanning every size category, optimized for every deployment scenario, available to anyone who wants to build with them.Whether &quot;distributed intelligence&quot; becomes the industry&#x27;s dominant paradigm or remains a compelling alternative serving a narrower market will determine not just Mistral&#x27;s fate, but the broader question of who controls the AI future — and whether that future will be open.For now, the race is on. And Mistral is betting it can win not by building the biggest model, but by building everywhere else.

أطلقت شركة Mistral AI عائلة Mistral 3، وهي مجموعة من 10 نماذج مفتوحة المصدر مصممة لتعمل على منصات متنوعة، بما في ذلك أجهزة الكمبيوتر المحمولة والطائرات بدون طيار وأجهزة الحافة. يمثل هذا الإطلاق خطوة مهمة في استراتيجية الشركة للتنافس مع الشركات الكبرى مثل OpenAI وGoogle من خلال توفير حلول ذكاء اصطناعي قابلة للوصول عبر تطبيقات مختلفة.

Mistral AI ha lanzado la familia Mistral 3, un conjunto de 10 modelos de IA de código abierto diseñados para diversas plataformas, incluidos portátiles, drones y dispositivos de borde. Este lanzamiento marca un paso significativo en la estrategia de la empresa para competir con grandes actores como OpenAI y Google al proporcionar soluciones de IA accesibles en diferentes aplicaciones.

Mistral AI a lancé la famille Mistral 3, une suite de 10 modèles d'IA open-source conçus pour diverses plateformes, y compris les ordinateurs portables, les drones et les dispositifs en périphérie. Ce lancement représente une étape significative dans la stratégie de l'entreprise pour rivaliser avec des acteurs majeurs comme OpenAI et Google en fournissant des solutions d'IA accessibles à travers différentes applications.

Mistral AI has launched the Mistral 3 family, a suite of 10 open-source AI models designed for various platforms, including laptops, drones, and edge devices. This release marks a significant step in the company's strategy to compete against major players like OpenAI and Google by providing accessible AI solutions across different applications.

Alibaba's AgentEvolver lifts model performance in tool use by ~30% using synthetic, auto-generated tasks

Was this article worth reading? Share it

Chattermate

Chattermate

Agentcloud