Term
Mistral Small 4
Mistral Small 4 is Mistral AI's open-weight MoE model from March 2026 — merging reasoning, vision and coding into one model, 119B parameters (6.5B active), Apache 2.0.
Mistral Small 4 — explained in more detail
Mistral AI released Small 4 on March 16, 2026, folding three previously separate products into a single model: Magistral (reasoning), Pixtral (vision) and Devstral (agentic coding). Architecture: Mixture-of-Experts with 128 experts, of which only 4 are activated per token — 119B total parameters, 6.5B active per inference. The context window is 256,000 tokens. Released under Apache 2.0 — commercial use, modification and self-hosting with no restrictions.
Example / Practical use
The killer feature is configurable reasoning depth: reasoning_effort="none" for fast chat (comparable to Small 3.2), reasoning_effort="high" for Magistral-level depth on complex tasks — one model, one deployment, switchable on the fly. The model handles text and images. Mistral reports 40% faster end-to-end response times and 3× more requests per second compared to Small 3. API pricing: $0.15 per 1M input tokens, $0.60 per 1M output tokens.
Entdecke mehr
Three Days with Claude Fable 5 — and Why Opus Stays My Daily Driver
Fable 5 fixed a bug in one shot that Opus failed at twice. I'm still not switching. My honest impressions from the trenches after three days.
GlossarMistral Large 3
Mistral Large 3 (December 2025) is Mistral's open-weight MoE flagship — 675B parameters, 256K-token context, natively multimodal, Apache 2.0.
LexikonComparing AI models — who builds what and how to choose
The major model families in 2026 at a glance. Who builds Claude, GPT, Gemini, Llama, Mistral, DeepSeek, Qwen — and which model to pick when.