Ministral 3 3B
Ministral 3 3B is Mistral's smallest current model: a 3.4B language model plus a 0.4B vision encoder under Apache 2.0. It keeps the family's 256K context window and function calling. Mistral says it fits in about 8 GB of VRAM at FP8, and an ONNX build is published.
Specifications
- Context window
- 256Kabout 197,000 words, or 390 pages
- Max output
- Not published
- Input price
- $0.10per 1M tokens
- Output price
- $0.10per 1M tokens
- Cached input
- Not published
- Weights
- OpenApache-2.0
- Parameters
- 3.8Bdense
- Input types
- Text, Images
- Output types
- Text
- Reasoning mode
- No
- Tool calling
- Yes
- Released
- Dec 2, 2025Current
Mistral API price (ministral-3b-2512); the docs list a single $0.1 per 1M rate, read as the same price for input and output.
What it costs in practice
| Workload | Monthly cost at list price |
|---|---|
| Support chatbot30,000 conversations a month, 2,000 input and 400 output tokens each | $7.20 |
| Document summaries2,000 documents a month, 12,000 input and 800 output tokens each | $2.56 |
| Coding agent1,000 tasks a month, 60,000 input and 6,000 output tokens each | $6.60 |
Standard rates, no caching or batch discounts. Run your own numbers.
Strengths
- Fits in about 8 GB of VRAM at FP8
- ONNX and GGUF builds available
- Apache 2.0 license
Trade-offs to weigh
- Small capacity limits accuracy on knowledge-heavy questions
- Best suited to narrow, well-defined tasks
Where it fits
Mistral AI positions Ministral 3 3B for:
Where you can use it
- Hugging Face
- Ollama
- API
Run it yourself
Released under Apache-2.0. Approximate memory needed just to load the weights:
| Precision | Memory for weights |
|---|---|
| 16-bit (BF16/FP16) | 8 GB |
| 8-bit (FP8/INT8) | 4 GB |
| 4-bit (Q4_K_M, about 4.85 bits) | 2 GB |
Add memory for the KV cache (it grows with context length and concurrent users) and 1-2 GB of runtime overhead. Size it precisely in the VRAM calculator.
Alternatives to consider
| Model | Tier | Context | Input / output per 1M | |
|---|---|---|---|---|
| Granite 4.2 8BIBM | Compact | 128K | $0.06 / $0.25 | Compare |
| NVIDIA Nemotron 3.5 Lightning 30B-A3BNVIDIA | Compact | 1M | $0.05 / $0.20 | Compare |
| gpt-oss-20bOpenAI | Compact | 128K | $0.075 / $0.30 | Compare |
| Gemma 4 26B A4BGoogle | Compact | 256K | $0.07 / $0.34 | Compare |
Frequently asked questions
How much does Ministral 3 3B cost?
Ministral 3 3B lists at $0.10 per million input tokens and $0.10 per million output tokens (checked October 10, 2026). Mistral API price (ministral-3b-2512); the docs list a single $0.1 per 1M rate, read as the same price for input and output. Confirm current rates on the provider's pricing page before you commit.
What is the context window of Ministral 3 3B?
Ministral 3 3B accepts up to 262,144 tokens of context (about 197,000 words, or 390 pages). Long prompts cost more and can be slower, so most apps send only what the task needs.
Is Ministral 3 3B open source?
Ministral 3 3B is an open-weight model released under the Apache-2.0 license. You can download the weights from Hugging Face (mistralai/Ministral-3-3B-Instruct-2512) and run or fine-tune them yourself, within the license terms. Open weights isn't the same as fully open source: training data and code may not be published.
Can I run Ministral 3 3B locally?
Yes, if you have enough memory. The weights alone need about 2 GB at 4-bit quantization and about 8 GB at 16-bit, plus room for the KV cache. Use the VRAM calculator on LLModel.com to check a specific GPU or Mac.
What is Ministral 3 3B best for?
Mistral AI positions Ministral 3 3B for on-device, low latency, high volume, fine-tuning. It accepts text, images as input, doesn't offer a separate reasoning mode and supports tool calling. Test it on a sample of your own tasks before you choose it for production.
Sources
- huggingface.co/mistralai/Ministral-3-3B-Instruct-2512
- docs.mistral.ai/models/ministral-3-3b-25-12
- huggingface.co/mistralai/Ministral-3-3B-Instruct-2512/raw/main/config.json
- ollama.com/library
Facts checked against these pages on October 10, 2026. Spot an error? Tell us.
Still choosing a model?
Send us your use case and constraints. We'll come back with a shortlist and an evaluation plan you can run on your own data.
- Free and vendor-neutral: no model provider pays for placement.
- Your details go only to partners you agree to.