Mistral AI

Mistral AI has made its Regional Endpoints generally available, letting customers choose EU or US inference, while opening its platform to third-party open models starting with Z.ai's GLM-5.2. The company also announced a coalition of European enterprises committing to long-term compute capacity, targeting up to 1GW by 2030.

Mistral AI has released Shieldstral-1.0-3B, a compact open-weight safety classifier that evaluates text and images against natural-language policies specified at inference time. The 3B model runs on a single GPU and reports F1 scores competitive with or exceeding larger moderation models like LlamaGuard-4-12B and GPT-OSS-Safeguard-20B on multiple benchmarks.

Mistral AI has released Leanstral 1.5, an open-source 119B parameter mixture-of-experts model designed specifically for Lean 4 proof assistance. The model features 128 experts with 4 active per token (6.5B activated parameters), a 256k token context window, and multimodal input capabilities.

Mistral AI released Leanstral 1.5, a free Apache-2.0 licensed model with 119B total parameters and 6B active parameters specialized for formal verification in Lean 4. The model achieves 100% on miniF2F benchmark, solves 587 of 672 PutnamBench problems at $4 per problem (versus $300+ for competitors), and reaches state-of-the-art 87% on FATE-H and 34% on FATE-X benchmarks.

Mistral AI released new enterprise connector features including workspace-level access controls, multi-account authentication for single connectors, and a debugging tool for Model Context Protocol (MCP) connections. The updates address production deployment challenges for AI agents accessing enterprise data systems.

Mistral AI released OCR 4, a compact document extraction model that returns bounding boxes, block classification, and inline confidence scores alongside text. The model supports 170 languages, scores 85.20 on OlmOCRBench, and is priced at $4 per 1,000 pages via API ($2 with batch discount) or $5 per 1,000 pages through Document AI.

Mistral AI has released Connectors in Studio, allowing developers to integrate custom MCP (Model Context Protocol) servers and built-in connectors via API/SDK. The release includes direct tool calling for deterministic workflows and human-in-the-loop approval flows for sensitive operations.

Mistral AI released Workflows in public preview, an orchestration layer built on Temporal's execution engine for running AI processes in production. The platform provides durable execution with state tracking, full observability through Studio, and single-line human-in-the-loop approval pauses. Organizations including ASML, ABANCA, and CMA-CGM are already using it to automate critical business processes.

Mistral AI has released Voxtral TTS, a 4B parameter text-to-speech model supporting 9 languages including English, French, German, Spanish, Dutch, Portuguese, Italian, Hindi, and Arabic. The model achieves 70ms latency for typical inputs and can clone voices from as little as 3 seconds of audio, priced at $0.016 per 1,000 characters.

Mistral AI has launched Forge, a platform that allows enterprises to train custom AI models on their proprietary data including codebases, compliance policies, and operational documentation. The system supports both dense and mixture-of-experts architectures with pre-training, post-training, and reinforcement learning capabilities.

Mistral AI has released Leanstral, a sparse 120B model with 6B active parameters designed specifically for the Lean 4 proof assistant. The model is available under Apache 2.0 license with free API access and achieves a 26.3 FLTEval score at pass@2, outperforming Claude Sonnet 4.6 while costing $36 versus $549.

Mistral AI released Vibe 2.0, a terminal-native coding agent powered by Devstral 2, adding custom subagents, multi-choice clarifications, and slash-command skills. Devstral 2 API pricing is now $0.40/M input tokens and $2.00/M output tokens, with a smaller variant at $0.10/$0.30 per million tokens.

Mistral AI's engineering team documented their investigation of a memory leak in vLLM that caused 400MB/minute memory growth during disaggregated serving with Mistral Medium 3.1. The leak, which only appeared with specific conditions including graph compilation and NIXL-based KV cache transfer, was eventually traced to mmap allocations outside the traditional heap that standard profiling tools couldn't detect.

Mistral AI released Mistral OCR 3, a document extraction model priced at $2 per 1,000 pages ($1 with Batch API discount). The model achieves a 74% overall win rate over its predecessor on forms, scanned documents, complex tables, and handwriting according to internal benchmarks.

Mistral has released Mistral 3, including Mistral Large 3—a sparse mixture-of-experts model with 41B active and 675B total parameters—and three Ministral 3 edge models (3B, 8B, 14B). All models are released under Apache 2.0 license with multimodal capabilities and are available today on multiple platforms.

Mistral AI launched AI Studio, a production platform designed to move enterprise AI systems from prototype to deployment. The platform includes three core components: Observability for tracking model performance, an Agent Runtime built on Temporal for durable execution, and an AI Registry for asset versioning and governance.

Mistral released 20+ MCP-powered connectors for Le Chat, integrating tools like Databricks, Snowflake, GitHub, Stripe, and Asana. The update includes a memory feature that saves user preferences across conversations, with one-click import from ChatGPT. All features are available on the free plan.

Mistral AI has published research showing that fine-tuning its Pixtral-12B vision language model on satellite imagery increases classification accuracy from 56% to 91% on the Aerial Image Dataset. Using Low-Rank Adaptation (LoRA) with 8,000 training samples across 30 scene categories, the company reduced hallucinations from 5% to 0.1% for under $10 in compute costs.

Mistral AI has released a major update to Le Chat, adding a Deep Research agent that generates structured research reports, a new voice input model called Voxtral, and Projects for organizing conversations. The update also includes multilingual reasoning powered by Mistral's Magistral model.

Mistral AI has released Codestral Embed, its first code-specialized embedding model, priced at $0.15 per million tokens. The model features an 8192-token context window and claims to outperform Voyage Code 3, Cohere Embed v4.0, and OpenAI's large embedding model on code retrieval benchmarks.