Mistral Large 4 vs Mistral Large 2.1

Mistral Large 4 is the better all-round choice, scoring 67/100 against Mistral Large 2.1 (38). It leads on capability, price, inputs & features and context window. The score weighs capability 50%, price 25%, inputs & features 15%, context window 10%.

Mistral Large 4 is cheaper at $0.70 input / $2.10 output per million tokens (official Mistral API price). Mistral Large 2.1 costs $2.00 input / $6.00 output per million tokens (median across 2 API providers). At a typical mix of three input tokens to one output token, that is $1.05 per million tokens for Mistral Large 4 versus $3.00 for Mistral Large 2.1 (2.9× as much).

Mistral Large 4 scores higher on the Capabilities Index (ECI): Mistral Large 4 150.6 (#57 of 165) and Mistral Large 2.1 128.6 (#142 of 165). Their confidence ranges do not overlap (147.4–153.5 vs 124.3–130.7), so the gap is a real one. On individual benchmarks: GPQA Diamond — Mistral Large 4 88.1%, Mistral Large 2.1 51.3%; OTIS Mock AIME 2024–2025 — Mistral Large 4 98.3%, Mistral Large 2.1 7.8%.

There are no published SWE-bench Verified results for Mistral Large 4 and Mistral Large 2.1 yet, so there is no like-for-like coding score. On overall capability, Mistral Large 4 leads, which tends to carry over to coding, but test on your own codebase. Both support tool calling for agent workflows.

Mistral Large 4 has the largest context window at 1,048,576 tokens, against 131,072 for Mistral Large 2.1. Maximum output per response: Mistral Large 4 up to 262,144, Mistral Large 2.1 up to 16,384 tokens.

Mistral Large 4 accepts text and images; Mistral Large 2.1 accepts text. Mistral Large 4 handles the widest range of inputs.

Mistral Large 2.1 publishes its weights and can be self-hosted; Mistral Large 4 is proprietary.

Mistral Large 4 is the newest, released Oct 6, 2026. Mistral Large 2.1 came out Nov 18, 2024. Knowledge cutoff: Mistral Large 2.1 Nov 2024.

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 Nov 2, 2026.