Mistral AI
Mistral AI is an artificial intelligence company founded in Paris, France, focused on the research and development of large language models. The company is known for its open-source strategy and has released model families including Mistral and Mixtral, covering both dense models and Mixture-of-Experts (MoE) architectures. It provides developers and enterprises with efficient, deployable AI foundation models and related API services. Mistral AI is a representative AI startup in Europe, dedicated to advancing open and high-performance language model development.
Core Facts
Mistral 7B was released in September 2023, has 7.3 billion parameters, and outperformed competing models with parameter counts of up to 13 billion and even 34 billion on multiple benchmarks.
Mistral Large 4 is the largest open-source model Mistral has released to date
Representative open-source AI models include Meta's Llama series, Mistral, and Stability AI.
Timeline (last 90 days)
Mistral AI released the Ministral 3 series model weights under the Apache 2.0 license, including two Instruct versions, 3B and 8B, both available on Hugging Face.
These three models were developed by Google DeepMind, the Alibaba Qwen team, and Mistral AI, respectively.
Mistral AI announced on social platforms that its model has entered space orbit, participating in the first AI satellite constellation project built by Orbitworks and Loft Orbital.
When core players in the ground AI ecosystem such as Mistral and NVIDIA begin to participate in orbital projects, it shows that on-orbit AI has moved from proof of concept to the engineering deployment stage.
NVIDIA AI Infrastructure provides AI computing infrastructure support for the AI satellite constellation project in which Mistral AI participates.
Mistral AI's official tweet uses the phrase "first-of-its-kind AI satellite constellation," indicating that the project is aimed at constellation-scale deployment rather than a single experimental satellite.
Mistral's Mixtral series adopts a sparse expert architecture, has a clear advantage in inference cost, and its efficiency-oriented technical路线 is naturally aligned with the power constraints of on-orbit deployment.
Mistral has offered a public preview of ML4 through Mistral Studio.
ML4's final license terms, deployment costs, and specific operational thresholds are currently not yet clear, and these details will only surface after the weights are actually released.
The company positions it as a frontier open-weights model for text and multimodal work, with application scenarios covering fields such as cybersecurity, finance, and law.
40 more timeline events
All Facts (20)
Mistral 7B was released in September 2023, has 7.3 billion parameters, and outperformed competing models with parameter counts of up to 13 billion and even 34 billion on multiple benchmarks.
80%VerifiedMistral Large 4 is the largest open-source model Mistral has released to date
75%VerifiedRepresentative open-source AI models include Meta's Llama series, Mistral, and Stability AI.
75%VerifiedMeta open-sourced the Llama series of models, Alibaba open-sourced the Qwen series, and Mistral AI has also adopted an open-source strategy.
75%VerifiedIn June 2024, Mistral AI completed a €600 million funding round, with its valuation jumping to approximately $6 billion, with investors including General Catalyst, Lightspeed Venture Partners, Samsung, NVIDIA, and Salesforce.
75%VerifiedMainstream open-source models such as Meta's Llama series, DeepSeek-R1, and Mistral mostly fall under the category of open-weight models.
70%VerifiedThe current open-source large model landscape has formed a multipolar competitive structure, including Meta's Llama, Alibaba's Qwen, Mistral AI, and DeepSeek.
65%VerifiedOpen-source large models such as Meta's LLaMA series, Mistral, and Falcon have formed an active open-source AI ecosystem
65%VerifiedMistral AI's Mixtral 8x7B adopts a MoE architecture with a total parameter count of approximately 47B, but only approximately 13B parameters are actually activated during each inference.
65%Verified开放权重模型仅发布训练完成的模型参数文件，而严格意义的开源AI要求公开训练代码、数据集与完整复现流程
65%VerifiedAmazon Bedrock provides access to foundation models from multiple vendors including Anthropic, Meta, Mistral AI, Cohere, and Stability AI through a unified API.
65%VerifiedMeta's LLaMA series, Google's Gemma, Mistral AI's Mistral/Mixtral, and Alibaba's Qwen series are all supported in Transformers.
65%VerifiedMistral has launched its new flagship model Mistral Large 4, codenamed 'Le Chonk'.
80%VerifiedMistral has released Mistral Large 4, a one-trillion-parameter, natively multimodal MoE model with approximately 49 billion active parameters, with open-source weights to be released at the end of the month.
75%UnverifiedMeta 的 LLaMA 系列、Mistral AI 的 Mistral/Mixtral、Google 的 Gemma 等开源模型都选择通过 Hugging Face 平台和 Transformers 格式进行发布
90%UnverifiedMistral AI的CEO Arthur Mensch曾在Google DeepMind担任研究科学家
90%UnverifiedMistral AI此前已发布Pixtral 12B和Pixtral Large等视觉语言模型
90%UnverifiedMistral AI推出了Le Chat——一款面向终端用户的AI助手产品，直接对标ChatGPT
90%UnverifiedMistral AI提供三种部署模式：云端API直接调用、通过Microsoft Azure/Google Cloud/AWS的托管部署，以及完全本地私有化部署
90%UnverifiedMistral AI的核心产品La Plateforme是一个云端API服务平台，提供模型调用、微调和Agent构建等能力
90%