Sponsored by Byond Boundrys Consulting - Empowering Ideas, Delivering Results
Mistral AI New Intermediate

Ministral 3 8B

Ministral 3 8B is a verified AI model profile covering official specifications, pricing or access, capabilities, practical use cases, strengths and limitations.

Multimodal Open-Weight ModelTextImage Freemium
In plain English

What is this model and why does it matter?

Ministral 3 8B is Mistral's efficient 8B text-and-vision model designed for edge and local deployment, with a 256K context window and open Apache-2.0 weights.

Local AIEdge assistantsDocument understandingVision Q&AStructured extraction
Model overview

Ministral 3 8B: features, use cases and important details

Ministral 3 8B is an AI model verified from first-party Mistral AI sources.

Ministral 3 8B verified specifications

Ministral 3 8B is Mistral’s efficient 8B text-and-vision model designed for edge and local deployment, with a 256K context window and open Apache-2.0 weights. Its verified context or usage limit is 256000 tokens.

Ministral 3 8B pricing and access

Ministral 3 8B costs $0.15 per 1M input tokens and $0.15 per 1M output tokens. Apache-2.0 weights are also available for self-hosting.

Ministral 3 8B best uses

Local AI, Edge assistants, Document understanding, Vision Q&A, Structured extraction.

Ministral 3 8B limitations

Long context still consumes memory, Vision errors require checking, Self-hosted operations remain your responsibility.

Get started

How to use this model

  1. Use the Mistral API or download official weights.
  2. Provide text or image input.
  3. Use function calling or structured outputs.
  4. Benchmark latency on your target edge hardware.
  5. Validate vision results.
Copy and try

Example prompts

  • Analyze this image and document.
  • Return these fields as JSON.
  • Run this assistant locally on private data.
Capabilities

What it can do

  • 8B parameters
  • 256K context
  • Text and vision
  • Function calling
  • Structured outputs
  • Apache 2.0
Best for

Practical use cases

  • Edge AI
  • Private assistants
  • Document intelligence
  • Visual Q&A
  • Extraction
Pricing

What does it cost?

Ministral 3 8B costs $0.15 per 1M input tokens and $0.15 per 1M output tokens. Apache-2.0 weights are also available for self-hosting.

Input$0.15 / 1M tokens
Output$0.15 / 1M tokens
Simple summaryHosted input and output are both $0.15/M tokens; self-hosting shifts cost to local or cloud GPU infrastructure.

What stands out

  • Open Apache license
  • Low API price
  • Multimodal
  • Edge-focused

Things to consider

  • Less capable than larger frontier models
  • Local speed depends on hardware
Limitations

Important restrictions and trade-offs

  • Long context still consumes memory
  • Vision errors require checking
  • Self-hosted operations remain your responsibility
SimplifyAITools verdict

Our editorial take

A strong high-demand compact model for users seeking local multimodal AI without moving to 14B or frontier-scale deployments.

References

Primary sources

  1. Open source 1 ↗
  2. Open source 2 ↗
  3. Open source 3 ↗
Keep exploring

Related AI models