Ministral 3 3B is Mistral AI's smallest edge model, using Cascade Distillation, tied embeddings, a frozen vision encoder, and a 256k context window.
Capabilities, design details, and architectural traits
Ministral 3 3B is the smallest and most efficient member of the Ministral 3 family. It is designed for edge deployment, combining compact size with native vision understanding rather than offering text-only capability.
| Trait | Detail |
|---|---|
| Pretraining method | Built using Cascade Distillation, a compute-efficient recipe that distills from a stronger, already post-trained teacher model (Mistral Medium 3.1) instead of training from scratch |
| Architecture | Uses tied embeddings, sharing the embedding and output layers, a design specific to the 3B size within the family since the 8B and 14B variants use separate layers for each |
| Vision encoder | Includes a vision encoder adapted from Mistral Small 3.1 Base, kept frozen, paired with a newly trained projection layer for image understanding |
| Variants | Released as base, instruct, and reasoning versions |
| Context window | Supports a 256k context window |
Ministral 3 3B is released under the Apache 2.0 license as open weights, and is built to run across diverse hardware, including local setups, rather than being restricted to large-scale cloud deployment.
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 | 131K | 1M Best Context Window | 1M Best Context Window |
| Pricing (per 1M tokens) | |||
| Input Pricing | $0.10 Best Input Pricing | $5 | $10 |
| Output Pricing | $0.10 Best Output Pricing | $25 | $50 |
| Modalities | |||
| Inputs | textimage | textimage | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 7.1 | 63.1 Best Intelligence Index | 62.1 |
| Coding Index | 4.8 | 78.0 Best Coding Index | 76.5 |
| Agentic Index | 1.6 | 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.
