Mistral Medium 3.5
● Excellent8.0
/ 10
Mistral Medium 3.5 is a dense 128B instruction-following model designed for agentic workflows, coding, and complex reasoning. Supports text, image, and file inputs with strong tool-use capabilities.
Specifications
| Attribute | Value |
|---|---|
| Lab | Mistral AI |
| Tags | Intelligent |
| Overall Score | 8.0/10 |
| Release Date | 2026-04 |
| Context Window | 262,144 tokens |
| Input Price / 1M | $1.50 |
| Output Price / 1M | $7.50 |
| Input Modalities | Text |
| Output Modalities | Text |
Strengths
- Dense 128B architecture for strong single-pass reasoning
- Purpose-built for agentic workflows and tool use
- Strong coding and software engineering performance
- Text, image, and file input support
- Competitive mid-tier pricing at $1.50/$7.50 per 1M tokens
Weaknesses
- No native reasoning/thinking mode
- 262K context window lags behind 1M+ flagships
- Higher output token cost than Mistral Large 3
Best For
- Agentic and tool-use workflows
- Coding and software engineering
- Complex multi-step reasoning tasks
- Cost-effective mid-tier inference
Sources & Further Reading
Related Models
Scores are aggregated from public benchmarks (MMLU, HumanEval, MATH, GSM8K, LMSYS) and normalized to a 1–10 scale. Methodology →