Ministral 3 8B is Mistral AI's edge model built via Cascade Distillation, offering base, instruct, and reasoning variants with vision and a 256K context.
Capabilities, design details, and architectural traits
Ministral 3 8B is a parameter-efficient dense language model from Mistral AI, built for compute and memory constrained applications such as edge and local deployment. It is part of the Ministral 3 family and is derived through a documented training recipe rather than standard pretraining alone.
| Trait | Detail |
|---|---|
| Training method | Cascade Distillation: an iterative pruning and continued training with distillation technique |
| Variants | Base, Instruct, and Reasoning, each with image understanding capabilities |
| Context window | Up to 256K tokens |
| Vision encoder | A 410M parameter ViT inherited from Mistral Small 3.1 Base, kept frozen, with a newly trained projection layer |
| License | Apache 2.0 |
The instruct variant is built to match or exceed the performance of comparable models while often producing an order of magnitude fewer output tokens, favoring efficiency over verbosity in real-world use.
The reasoning variant spends more inference time working through a problem before answering, aiming for state-of-the-art accuracy within its weight class when accuracy matters more than speed.
Independent evaluations · Artificial Analysis
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Mistral | Anthropic | Anthropic |
| Release Date | December 2, 2025 | July 24, 2026 | June 9, 2026 |
| Knowledge Cutoff | - | May 2026 | - |
| Context & Limits | |||
| Context Window | 262K | 1M Best Context Window | 1M Best Context Window |
| Pricing (per 1M tokens) | |||
| Input Pricing | $0.15 Best Input Pricing | $5 | $10 |
| Output Pricing | $0.15 Best Output Pricing | $25 | $50 |
| Modalities | |||
| Inputs | textimage | textimage | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 9.0 | 63.1 Best Intelligence Index | 62.1 |
| Coding Index | 9.7 | 78.0 Best Coding Index | 76.5 |
| Agentic Index | 1.2 | 59.2 Best Agentic Index | 56.6 |
Graduate-level reasoning and expert Q&A evaluation.
Extremely difficult logical reasoning and knowledge.
Logical reasoning over long context windows.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
