Introducing Mistral 3 | Mistral AI

Today, we announce Mistral 3, the subsequent generation of Mistral models. Mistral 3 includes three state-of-the-art small, dense models (14B, 8B, and 3B) and Mistral Large 3 – our most capable model so far – a sparse mixture-of-experts trained with 41B energetic and 675B total parameters. All models are released under the Apache 2.0 license. Open-sourcing our models in a wide range of compressed formats empowers the developer community and puts AI in people’s hands through distributed intelligence.

The Ministral models represent the very best performance-to-cost ratio of their category. At the identical time, Mistral Large 3 joins the ranks of frontier instruction-fine-tuned open-source models.

Mistral Large 3: A state-of-the-art open model

Mistral Large 3 is probably the greatest permissive open weight models on this planet, trained from scratch on 3000 of NVIDIA’s H200 GPUs. Mistral Large 3 is Mistral’s first mixture-of-experts model because the seminal Mixtral series, and represents a considerable step forward in pretraining at Mistral. After post-training, the model achieves parity with the very best instruction-tuned open-weight models in the marketplace on general prompts, while also demonstrating image understanding and best-in-class performance on multilingual conversations (i.e., non-English/Chinese).

Mistral Large 3 debuts at #2 within the OSS non-reasoning models category (#6 overall amongst OSS models overall) on the LMArena leaderboard.

We release each the bottom and instruction fine-tuned versions of Mistral Large 3 under the Apache 2.0 license, providing a powerful foundation for further customization across the enterprise and developer communities. A reasoning version is coming soon! 

Mistral, NVIDIA, vLLM & Red Hat join forces to deliver faster, more accessible Mistral 3

Working together with vLLM and Red Hat, Mistral Large 3 could be very accessible to the open-source community. We’re releasing a checkpoint in NVFP4 format, built with llm-compressor. This optimized checkpoint allows you to run Mistral Large 3 efficiently on Blackwell NVL72 systems and on a single 8×A100 or 8×H100 node using vLLM.

Delivering advanced open-source AI models requires broad optimization, achieved through a partnership with NVIDIA. All our latest Mistral 3 models, from Large 3 to Ministral 3, were trained on NVIDIA Hopper GPUs to tap high-bandwidth HBM3e memory for frontier-scale workloads. NVIDIA’s extreme co-design approach brings hardware, software, and models together. NVIDIA engineers enabled efficient inference support for TensorRT-LLM and SGLang for the entire Mistral 3 family, for efficient low-precision execution.

For Large 3’s sparse MoE architecture, NVIDIA integrated state-of-the-art Blackwell attention and MoE kernels, added support for prefill/decode disaggregated serving, and collaborated with Mistral on speculative decoding, enabling developers to efficiently serve long-context, high-throughput workloads on GB200 NVL72 and beyond. On the sting, delivers optimized deployments of the Ministral models on DGX Spark, RTX PCs and laptops, and Jetson devices, giving developers a consistent, high-performance path to run these open models from data center to robot.

We’re very thankful for the collaboration and wish to thank vLLM, Red Hat, and NVIDIA particularly.

Ministral 3: State-of-the-art intelligence at the sting

For edge and native use cases, we release the Ministral 3 series, available in three model sizes: 3B, 8B, and 14B parameters. Moreover, for every model size, we release base, instruct, and reasoning variants to the community, each with image understanding capabilities, all under the Apache 2.0 license. When married with the models’ native multimodal and multilingual capabilities, the Ministral 3 family offers a model for all enterprise or developer needs.

Moreover, Ministral 3 achieves the very best cost-to-performance ratio of any OSS model. In real-world use cases, each the variety of generated tokens and model size matter equally. The Ministral instruct models match or exceed the performance of comparable models while often producing an order of magnitude fewer tokens. 

For settings where accuracy is the one concern, the Ministral reasoning variants can think longer to supply state-of-the-art accuracy amongst their weight class – for example 85% on AIME ‘25 with our 14B variant.

Available Today

Mistral 3 is obtainable today on Mistral AI Studio, Amazon Bedrock, Azure Foundry, Hugging Face (Large 3 & Ministral), Modal, IBM WatsonX, OpenRouter, Fireworks, Unsloth AI, and Together AI. As well as, coming soon on NVIDIA NIM and AWS SageMaker.

Yet one more thing… customization with Mistral AI

For organizations in search of tailored AI solutions, Mistral AI offers custom model training services to fine-tune or fully adapt our models to your specific needs. Whether optimizing for domain-specific tasks, enhancing performance on proprietary datasets, or deploying models in unique environments, our team collaborates with you to construct AI systems that align together with your goals. For enterprise-grade deployments, custom training ensures your AI solution delivers maximum impact securely, efficiently, and at scale.

Start with Mistral 3

The longer term of AI is open. Mistral 3 redefines what’s possible with a family of models built for frontier intelligence, multimodal flexibility, and unmatched customization. Whether you’re deploying edge-optimized solutions with Ministral 3 or pushing the boundaries of reasoning with Mistral Large 3, this release puts state-of-the-art AI directly into your hands.

Why Mistral 3?

• Frontier performance, open access: Achieve closed-source-level results with the transparency and control of open-source models.

• Multimodal and multilingual: Construct applications that understand text, images, and sophisticated logic across 40+ native languages.

• Scalable efficiency: From 3B to 675B energetic parameters, select the model that matches your needs, from edge devices to enterprise workflows.

• Agentic and adaptable: Deploy for coding, creative collaboration, document evaluation, or tool-use workflows with precision.

Next Steps

1. Explore the model documentation: 

Science has all the time thrived on openness and shared discovery. As pioneering French scientist and two-time Nobel laureate Marie Skłodowska-Curie once said, “Nothing in life is to be feared, it is barely to be understood. Now could be the time to grasp more, in order that we may fear less.” 

This philosophy drives our mission at Mistral AI. We imagine that the long run of AI needs to be built on transparency, accessibility, and collective progress. With this release, we invite the world to explore, construct, and innovate with us, unlocking latest possibilities in reasoning, efficiency, and real-world applications.

Together, let’s turn understanding into motion.