Mistral Small 4 (v26.03) NEW
Mistral Small 4 (v26.03) is an API model from Mistral AI. It’s positioned for hard reasoning and planning—work that benefits from iteration, not just one-shot answers.
Intelligence
Medium
Speed
Good
171 tok/s output
Cost
Low
$0.15 in / $0.60 out
Context
256K
Up to 256,000 tokens
How are Intelligence, Speed & Cost bucketed?
Intelligence and Speed buckets are percentile ranks on
Artificial Analysis. Cost buckets are fixed dollar
thresholds keyed off output-token price ($/M out).
Intelligence
- Top 1%≤ 1%
- Top 5%≤ 5%
- Top 10%≤ 10%
- Good≤ 25%
- Medium≤ 50%
- Below avg> 50%
Speed
- Top 1%≥ 345 tok/s
- Top 5%≥ 237 tok/s
- Top 10%≥ 196 tok/s
- Good≥ 146 tok/s
- Medium≥ 90 tok/s
- Slow< 90 tok/s
Cost
- Freeopen weights · self-host
- Low< $1 / M out
- Moderate$1–5 / M out
- High≥ $5 / M out
Core Capabilities
Long Documents
Handles entire codebases, books, and multi-doc RAG.
Reasoning
Solves complex math, logic, and planning tasks.
Context Window
256k tokens
≈ 197+ pages
4k Chat 聊天
32k Long docs 长文档
128k Books 整本书
400k Multi-doc 多文档
1M Codebase 整个代码库
10M
256k
Availability
API
Available
Product / App
Not available
Open Source
Not released
Enterprise
Contact sales
Pricing Model
Free / self-host
Open weights — pay only for compute.
Self-host Capability / Performance
Where this model sits relative to the middle 60% of models in the tree. All scores are 0–10 (higher is better).
Lower 20% Upper 80% This model
Context / memory
Context window size · log-scaled
6.0
9.0
7.0
Lower 20% 20th percentile — 20% of models score below this This model Where the current model lands Upper 80% 80th percentile — only 20% of models score above this
Percentile boundaries are computed across every model in the tree that reports the underlying benchmark for each capability.
What it feels like
- Small tier — laptop-class deployment.
- Mistral's most capable model at release — on par with GPT-4o, Claude 3 Opus, Llama 3 405B at 1/3 the parameter count
- 84.0% MMLU on the pretrained variant — set a new performance/cost Pareto point for open weights
- Single-node inference: 123B dense fits on one GPU server — no MoE-routing complexity
Best use cases
- Multilingual production apps — French/German/Spanish/Italian/Arabic/Chinese/Japanese first-class
- Function-calling agents and business integrations needing parallel tool calls
- 80+ programming languages including Python/Java/C++/JavaScript/Bash
Tools to try
Not ideal for
- Frontier reasoning leaderboards — newer Opus / GPT-5 / DeepSeek R1 lead by significant margins
- Free-to-use commercial deployments (research license requires payment for production)