Mistral AI

Ministral 3 3B

Generally availableOpen weights131K context

Summary

Ministral 3 3B is an open-weight language model from Mistral AI, released on 2 Dec 2025. It accepts text, images and PDFs and generates text, with a 131K-token context window and up to 131K output tokens per response. As of 18 Nov 2026, the lowest listed API price is $0.10 per million input tokens and $0.10 per million output tokens (Mistral API), across 3 providers we track.

Compact open vision-language model for edge deployment, instruction following, and tool use

Last verified 18 Nov 2026. Every figure on this page links to its source.

Specifications

DeveloperMistral AI
Release date2 Dec 2025 [models.dev]
StatusGenerally available [models.dev]
WeightsOpen weights [Hugging Face]
Licenseapache-2.0 [Hugging Face]
Parameters3.8B [Hugging Face]
Context window131K (131,072 tokens) [models.dev]
Max output131K tokens [models.dev]
InputText, images and PDFs [models.dev]
OutputText [models.dev]
Reasoning modeNo [models.dev]
Tool callingYes [models.dev]
Structured outputYes [models.dev]
Hugging Facemistralai/Ministral-3-3B-Instruct-2512 [Hugging Face]

Ministral 3 3B API pricing by provider

USD per million tokens. Sorted by input price.

ProviderInputOutputCached inputContextSource
Mistral API
ministral-3b-2512
$0.10$0.10$0.01131Kmodels.dev, 18 Nov 2026
Amazon Bedrock
mistral.ministral-3-3b-instruct
$0.10$0.10–256Kmodels.dev, 18 Nov 2026
OpenRouter
mistralai/ministral-3b-2512
$0.10$0.10$0.01131KOpenRouter, 18 Nov 2026

Running Ministral 3 3B locally

Estimated memory, from the parameter count and the architecture in config.json on Hugging Face. Includes about 0.8 GB of runtime overhead.

QuantizationWeightsTotal, 8K contextTotal, 32K context
Q4_K_M (4-bit)2.3 GB4.0 GB6.6 GB
Q8_0 (8-bit)4.1 GB5.8 GB8.4 GB
FP16 / BF16 (16-bit)7.7 GB9.4 GB12 GB

At 4-bit with an 8K conversation, Ministral 3 3B needs about 4.0 GB, so it fits on a graphics card with 8 GB or more, such as the NVIDIA GeForce RTX 4060 8 GB. Check your own hardware, or see Running local LLMs on an 8 GB GPU on our sister site AIPCs.

Frequently asked questions

What is the context window of Ministral 3 3B?

Ministral 3 3B has a context window of 131,072 tokens (131K) and can generate up to 131,072 tokens in a single response, according to models.dev.

How much does the Ministral 3 3B API cost?

Through Mistral API, Ministral 3 3B costs $0.10 per million input tokens and $0.10 per million output tokens. Prices last verified 18 Nov 2026.

Is Ministral 3 3B open source?

Ministral 3 3B is an open-weight model: its weights are published on Hugging Face as mistralai/Ministral-3-3B-Instruct-2512 under the apache-2.0 license. Check the license terms before commercial use.

When was Ministral 3 3B released?

Mistral AI released Ministral 3 3B on 2 Dec 2025, according to models.dev.

Which providers offer Ministral 3 3B?

We track Ministral 3 3B on 3 providers: Mistral API, Amazon Bedrock, OpenRouter.

How much VRAM does Ministral 3 3B need?

About 4.0 GB at 4-bit (Q4_K_M) with an 8K-token conversation, and about 5.8 GB at 8-bit. The 4-bit model fits on a graphics card with 8 GB or more, such as the NVIDIA GeForce RTX 4060 8 GB. These are estimates from the model's published parameter count and architecture.

Similar models from other labs

Comparable on output type, weights, release date, price and context window. Prices are the lowest listed, USD per million tokens.

ModelLabReleasedContextInput $/1MOutput $/1MWeights
GPT OSS Safeguard 20BOpenAI29 Oct 2025128K$0.07$0.20Open
gpt-oss-120bOpenAI5 Aug 2025131K$0.04$0.17Open
Granite 4.0 H MicroIBM2 Oct 2025131K$0.02$0.11Open
AutoGLM Phone 9B MultilingualZhipu AI (Z.ai)10 Dec 202566K$0.04$0.14Open
Nvidia Nemotron Nano 9B V2NVIDIA18 Aug 2025131K$0.06$0.23Open
Baichuan M2 32BBaichuan13 Aug 2025131K$0.07$0.07Open

Change history

  • Max output changed from 262K to 131K
  • Input modalities changed from text, image to text, image, pdf
  • Context window changed from 256K to 131K
  • Max output changed from 8K to 262K
  • Now listed on Mistral API

Sources