Mistral Small 4
One Apache-2.0 model replacing Magistral, Pixtral and Devstral: 119B MoE with 6B active, so it fits one H100 at 4-bit and runs at ~165 tok/s. $0.15/$0.60 is hard to beat for a self-hostable multimodal model, but it is not clever.
Key facts
- Score
- 42 / 100, rank #24Evals19Price77People48Fit85
- Price
- $0.15 in · $0.60 out per 1M tokens (Mistral Small 4 announcement)
- Blended price
- $0.26 per 1M tokens (3 input : 1 output)
- Context window
- 256K
- Open weights
- Yes
- Released
- Mar 16, 2026
- Best for
- Self-hosted multimodal chat on one GPU, Fine-tuning base for EU-regulated teams, Cheap fast tasks with a reasoning_effort dial
Mistral Small 4 costs $0.15 per million input tokens and $0.60 per million output tokens on Mistral AI's list price (Mistral Small 4 announcement, read Oct 11, 2026).
Independent evals
| Benchmark | Result | Who ran it | Read |
|---|---|---|---|
| Artificial Analysis Intelligence Index (reasoning) | 11 | Mistral Small 4 - Artificial Analysis | Oct 10, 2026 |
- Artificial Analysis output speed (tokens/s): 165.4 (Mistral Small 4 - Artificial Analysis)
What practitioners say
Modest HN reception (126 points). Fans liked the economics: '$0.60/1M output is a steal' versus Qwen models that 'waste tokens on reasoning', and that ~120B 'fits onto a single H100 with 4 bit quant'. Others were blunt: 'okay, nothing exceptional', 'I'd use it for some basic tasks but not actual complex tasks', and one tester called it 'worse than glm air 4.5'. A later comment found the 4-bit quant roughly comparable to Qwen 3.6 27B.
Caveats
- AA 11; HN testers rate it 'okay, nothing exceptional' and 'not the best' for complex agentic work.
- Not listed on Arena; limited independent evaluation.
- Mistral Medium 3.5 (Modified MIT) is the step up if you need more intelligence from the same vendor.
Questions people ask
How much does Mistral Small 4 cost?
Mistral Small 4 costs $0.15 per million input tokens and $0.60 per million output tokens on Mistral AI's list price (Mistral Small 4 announcement, read Oct 11, 2026).
How good is Mistral Small 4?
Mistral Small 4 ranks #24 of 26 on the Models at Work leaderboard with a score of 42 out of 100 (October 2026 edition). One Apache-2.0 model replacing Magistral, Pixtral and Devstral: 119B MoE with 6B active, so it fits one H100 at 4-bit and runs at ~165 tok/s. $0.15/$0.60 is hard to beat for a self-hostable multimodal model, but it is not clever.
What is Mistral Small 4 best for?
Mistral Small 4 is best for self-hosted multimodal chat on one gpu, fine-tuning base for eu-regulated teams, cheap fast tasks with a reasoning_effort dial.
What are Mistral Small 4's benchmark scores?
Artificial Analysis Intelligence Index (reasoning): 11 (Mistral Small 4 - Artificial Analysis).
What is Mistral Small 4's context window?
Mistral Small 4 has a 256K token context window according to Mistral AI.
Is Mistral Small 4 open weights?
Yes. Mistral Small 4 is released with open weights, so it can be self-hosted.
Compare Mistral Small 4
Methodology: Score is 0 to 100 and computed, not typed: 55% independent evals (Artificial Analysis, Arena, Scale SEAL, Epoch, each scaled against its natural floor and the best score in this table, then averaged), 15% blended price on a fixed log scale ($0.05 per million is 100, $60 is 0), 15% practitioner sentiment from Signals and community threads, 15% operational fit (context window, open weights). A missing component drops out and the rest are reweighted. A model with no independent eval yet is provisional and ranks below every measured one. Within three points is a tie. Vendor-published figures are labelled and never counted.