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

Mistral Large 3

Mistral Large 3 is Mistral's 675B/41B active multimodal flagship with 256K context, Apache 2.0 weights and current API pricing.

Text GenerationText Paid
In plain English

What is this model and why does it matter?

Mistral Large 3 is Mistral's open-weight 675B-total, 41B-active multimodal flagship with 256K context.

Frontier open-weight deploymentMultimodal assistantsComplex agentsCodingDocument analysisEnterprise reasoning
Model overview

Mistral Large 3: features, use cases and important details

Mistral Large 3 is the current model represented by this generic Mistral Large page.

Verified model facts

Mistral documents a granular MoE architecture with 675B total and 41B active parameters, 256K context, multimodal input, Apache 2.0 licensing, function calling and structured outputs.

Current status

The model is active in the Mistral API as mistral-large-2512 and the weights are available.

Best fit

It is best for complex enterprise agents, document analysis, coding and teams that want frontier-scale open weights.

Limitations

Its size makes self-hosting demanding, and Mistral Small 4 can be a better cost/performance choice for many workloads.

Get started

How to use this model

  1. Use model mistral-large-2512 in the Mistral API or deploy the official weights.
  2. Provide text and optional image input.
  3. Use function calling or structured outputs when needed.
  4. Monitor token cost for long 256K prompts.
  5. Evaluate self-hosting only with suitable large-scale infrastructure.
Copy and try

Example prompts

  • Analyze these documents and images and produce an executive recommendation.
  • Use the available tools to complete this complex workflow.
  • Review this architecture and return a structured risk assessment.
Capabilities

What it can do

  • 256K context
  • Multimodal input
  • 41B active MoE
  • Function calling
  • Structured outputs
  • Open weights
Best for

Practical use cases

  • Enterprise agents
  • Complex document analysis
  • Coding
  • Research
  • Large-scale self-hosting
Pricing

What does it cost?

$0.50 per 1M input tokens and $1.50 per 1M output tokens on Mistral's current API.

Input$0.50 / 1M tokens
Output$1.50 / 1M tokens
Simple summaryCurrent API pricing is $0.50/M input and $1.50/M output; self-hosting the 675B model requires very large infrastructure.

What stands out

  • Apache 2.0
  • Frontier-scale open weights
  • 256K context
  • Multimodal
  • Current API

Things to consider

  • 675B total parameters
  • Heavy self-hosting requirements
  • More expensive than Mistral Small 4
Limitations

Important restrictions and trade-offs

  • Can hallucinate
  • Long-context usage can be costly
  • Full self-hosting requires specialized distributed infrastructure
SimplifyAITools verdict

Our editorial take

A strong current Mistral flagship for organizations needing frontier-scale open weights and multimodality, while Small 4 is usually the more economical choice.

References

Primary sources

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

Related AI models