Mistral LLM
2 Reviews and Ratings
What is Mistral LLM?
Mistral LLM is Mistral AI's family of large language models, offered as open-source weights free under an Apache 2.0 license and as commercial models served through the Mistral AI API. The models are aimed at business use cases and leverage private data and usage feedback. The original Mistral 7B transformer model was built for fast deployment and easy customization, though Mistral 7B and the Mixtral 8x7B/8x22B models were retired on March 30, 2025; their documented replacements are Ministral 3 8B and Mistral Small 4.
Flagship: Mistral Large 4. Mistral Large 4 (API name mistral-large-4) is the current flagship, generally available since November 5, 2026, following a public preview announced October 6, 2026. It is an open-weight, natively multimodal, general-purpose model with a granular Mixture-of-Experts (MoE) architecture: 1.05T total parameters with 52B active, plus a 1.6B-parameter vision encoder, and a 1M-token context window accepting text and image input. It was trained from scratch in Mistral's own European datacenters, with multilingual training data spanning more than 160 languages. API pricing is $0.7 per 1M input tokens, $0.07 per 1M cached input tokens, and $2.1 per 1M output tokens.
Model series and evolution. The current lineup spans three series:
- Large — Mistral Large 4 (
mistral-large-4, v26.10): granular MoE, 1.05T total / 52B active parameters plus a 1.6B vision encoder, 1M-token context, text and image input; open-weight. - Medium — Mistral Medium 3.5 (
mistral-medium-3-5, v26.04): 128B merged dense model with multimodal input and configurable reasoning effort; commercial license, deprecated October 31, 2026 in favor of Large 4. - Small — Mistral Small 4 (
mistral-small-4-0-26-03, v26.03): hybrid MoE with 119B total / 6.5B active parameters, unifying instruct, reasoning, and coding in one multimodal model; Apache 2.0.
The open-weight line evolved from Mistral 7B (September 2023, dense) through the Mixtral 8x7B and 8x22B sparse-MoE models and Mistral Nemo 12B, and the specialist lines (Codestral for code, Devstral, Magistral for reasoning, Pixtral for vision), into the current Ministral 3 series — 3B, 8B, and 14B text-and-vision models under Apache 2.0 — and Mistral Small 4. The commercial series progressed from Large 1.0 (February 2024) through Large 2.0/2.1 (2024) and Large 3 (December 2025) to Large 4; from Medium 1.0 (December 2023) through Medium 3 (May 2025) to Medium 3.5 (April 2026); and from Small 1.0 (February 2024) through the Small 3.x releases to Small 4.
Mistral AI API capabilities. The Mistral AI APIs support LLM applications through:
- Text generation with streaming, displaying partial model results in real time
- Embeddings for RAG, representing the meaning of text as a list of numbers
- Function calling, connecting Mistral models to external tools
- JSON mode, setting the response format to
json_object - Guardrailing, enforcing policies at the system level of Mistral models
Categories & Use Cases
Technical Details
| Mobile Application | No |
|---|