Mistral AI
✓ VerifiedEuropean provider of open-weights and hosted models.
At a glance
Overview
Mistral AI is a French artificial-intelligence company founded in April 2023 by alumni of Google DeepMind and Meta AI. Within months of founding, the company raised a €105 million seed round at a €240 million valuation, and went on to reach a US$6 billion valuation by mid-2024 — making it Europe's most valuable AI company.
Open-weights philosophy
Mistral has pursued a hybrid strategy: releasing strong open-weights models alongside commercial closed offerings.
- Mistral 7B (Sep 2023) — Apache-2.0 model that outperformed Llama 2 13B
- Mixtral 8x7B (Dec 2023) — sparse mixture-of-experts open release
- Mistral Large (Feb 2024) — closed flagship, available via Microsoft Azure
- Codestral (May 2024) — code-specialised model
- Devstral (2025) — frontier coding model
Le Chat
Mistral operates Le Chat, a free consumer chatbot competing with ChatGPT, marketed in part on its European data residency and lower pricing.
EU positioning
Mistral has been a vocal advocate for European AI sovereignty during the development of the EU AI Act, arguing for lighter-touch regulation of foundation-model providers.
Latest activity
Full history →- 1
Mistral Large pricing update
Mistral Large: input $2.00 → $0.50 /1M tokens, output $6.00 → $1.50 /1M tokens
Mistral Large - 1
Codestral pricing update
Codestral: input $0.30 → $1.00 /1M tokens, output $0.90 → $3.00 /1M tokens
Codestral - Capability
Mistral Large updated — improved tool calling
Mistral Large receives improved function-calling reliability and larger context in the November 2024 update.
Mistral Large - Release
Codestral released
Mistral launches Codestral, a 22B code-specialized model with 256k context and FIM support.
Codestral
Releases timeline
Showing 11 most recent- Codestral 2508
Specialized code generation model.
- Devstral
Agentic coding model for software development.
- Mistral Large 3
Top-tier reasoning and coding model.
- Mistral Medium 3
Balanced performance and cost.
- Mistral Small 3.2
Fast and affordable.
- Mistral Large
Mistral's flagship commercial model with tool calling and structured outputs.
- Mistral Small
Efficient open-weights mid-sized model from Mistral.
- Mistral NeMo
12B open-weights model built with NVIDIA, with 128k context.
- Codestral
Mistral's code-specialized model with long context.
- Mixtral 8x7B
Open-weights mixture-of-experts model with GPT-3.5 class performance.
- Mistral 7B
Compact open-weights model that outperforms Llama 2 13B on many benchmarks.
Active models
- CodestralActive
Mistral's code-specialized model with long context.
32K ctx - Codestral 2508Active
Specialized code generation model.
256K ctx - DevstralActive
Agentic coding model for software development.
256K ctx - Mistral 7BActive
Compact open-weights model that outperforms Llama 2 13B on many benchmarks.
32K ctx - Mistral LargeActive
Mistral's flagship commercial model with tool calling and structured outputs.
262K ctx - Mistral Large 3Active
Top-tier reasoning and coding model.
262K ctx - Mistral Medium 3Active
Balanced performance and cost.
- Mistral NeMoActive
12B open-weights model built with NVIDIA, with 128k context.
128K ctx - Mistral SmallActive
Efficient open-weights mid-sized model from Mistral.
32K ctx - Mistral Small 3.2Active
Fast and affordable.
- Mixtral 8x7BActive
Open-weights mixture-of-experts model with GPT-3.5 class performance.
32K ctx
Community ratings
Rate Mistral AI
Sign in to rate and review.
Comments
Sign in to leave a comment.