← All modelsbalanced
Mistral: Mistral Large 4
Mistral
Capabilities
VisionToolsStructured OutputExtended ThinkingLong ContextCode
Strengths
- Exceptional 1M-token context window for long-document tasks
- Strong performance on math (0.93) and instruction-following
- Multimodal vision capabilities for image understanding
- Extended thinking and structured output support
Weaknesses
- Moderate speed for real-time applications
- Limited transparency on training data and methodology
- Closed-source weights limit self-hosting and privacy control
- Vision fitness slightly lower than text-only performance
Pricing
Input / 1M tokens
$0.6799999999999999
Output / 1M tokens
$2.09
Context window
1,048.576k
Transparency
Open weights
0.0 / 10
Open training data
0.0 / 10
Open methodology
3.5 / 10
Licence openness
4.0 / 10
Provider disclosure
4.5 / 10
FMTI company score
N/A
Composite:5.5 / 10
Mistral Large 4 is closed-source with limited public methodology disclosure. Mistral publishes some technical details but significantly less than Meta or DeepSeek. No open training data. Licence terms are proprietary.
Sustainability
Inference energy
N/A
Training footprint
N/A
Provider infrastructure
5.5 / 10
Composite:5.0 / 10
Mistral's sustainability commitments are not well-documented publicly. Provider infrastructure score reflects uncertain green energy usage. No detailed inference or training energy data available.
Task Fitness
question answering
75%
code
77%
mathematics
93%
business communication
85%
vision
75%
analysis
77%
extraction
85%
long-form generation
77%
research
80%
reasoning
76%
summarisation
82%
translation
78%
conversation
84%