Plain-English Summary
Mistral’s large open-weight multimodal model for enterprise assistants, agents, and long-context work.
Mistral Large 3
Mistral’s large open-weight multimodal model for enterprise assistants, agents, and long-context work.
Mistral Large
Multimodal LLM
GA / open weights
Open weights: Yes
API: Yes and self-hosted
Technical Notes
675B-total, 41B-active multimodal MoE with 256K context, text-and-image input, tool use, and structured output.
Best For
Enterprise RAG, agents, coding, multilingual multimodal applications
Watch Out For
Large self-hosting footprint despite sparse activation.
Model Specs
- Maker
- Mistral AI
- Country
- France
- Family
- Mistral Large
- Type
- Multimodal LLM
- Release date
- 2025-12-02
- Date confidence
- Exact
- Status
- GA / open weights
- Context window
- 262,144 tokens
- Max output
- Not listed
- Parameters
- 675B
- Active parameters
- 41B
- Architecture
- Mixture-of-experts multimodal transformer
- Reasoning
- General reasoning
- Tool calling
- Yes
- Structured output
- Yes
- API available
- Yes and self-hosted
- Open weights
- Yes
- Self-hostable
- Yes
- License
- Apache 2.0
- Inputs
- Text, image
- Outputs
- Text
- Approx. price
- $0.50 input / $1.50 output per 1M tokens
- How to access
- Provider website
- Model / API ID
- mistral-large-3-25-12
- Knowledge cutoff
- Not publicly disclosed
- Fine-tuning
- Fine-tuning, self-hosting, tools
- Completeness
- 95 %
- Verified
- Jul 18, 2026
- Snapshot date
- Jul 18, 2026
- Source
- Official source