Magistral Small vs Mistral Medium 3

It is close. Our weighted score puts them within 3 points (Mistral Medium 3 53/100, Magistral Small 50/100), so choose by what matters most for your work: Mistral Medium 3 for raw capability and Magistral Small on price. The score weighs capability 50%, price 25%, inputs & features 15%, context window 10%.

Magistral Small is cheaper at $0.50 input / $1.50 output per million tokens (median across 1 API provider). Mistral Medium 3 costs $0.40 input / $2.00 output per million tokens (official Mistral API price). At a typical mix of three input tokens to one output token, that is $0.75 per million tokens for Magistral Small versus $0.80 for Mistral Medium 3 (1.1× as much).

Mistral Medium 3 scores higher on the Capabilities Index (ECI): Mistral Medium 3 134.1 (#117 of 148) and Magistral Small 133.2 (#120 of 148). The confidence ranges of the top two overlap (130.5–135.6 vs 128.9–134.4), so treat the gap as small. On individual benchmarks: GPQA Diamond — Mistral Medium 3 59.5%, Magistral Small 56.1%; OTIS Mock AIME 2024–2025 — Mistral Medium 3 32.2%, Magistral Small 30.0%.

There are no published SWE-bench Verified results for Magistral Small and Mistral Medium 3 yet, so there is no like-for-like coding score. On overall capability, Mistral Medium 3 leads, which tends to carry over to coding, but test on your own codebase. Both support tool calling for agent workflows.

Magistral Small and Mistral Medium 3 share the same 131,072-token context window. Maximum output per response: Magistral Small up to 8,192, Mistral Medium 3 up to 131,072 tokens.

Magistral Small accepts text; Mistral Medium 3 accepts text and images. Mistral Medium 3 handles the widest range of inputs.

Magistral Small publishes its weights (Apache 2.0) and can be self-hosted; Mistral Medium 3 is proprietary.

Magistral Small is the newest, released Jun 10, 2025. Mistral Medium 3 came out May 7, 2025. Knowledge cutoff: Mistral Medium 3 May 2025.

Each model gets a 0–100 score on capability (50%, independent benchmark results); price (25%, blended price per million tokens (3 input : 1 output), log scale); inputs & features (15%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (10%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Oct 4, 2026.