Mistral Large 2
Mistral AI
Mistral AI's flagship model with 123B parameters designed for single-node inference at large throughput. Released July 2024 with breakthrough code generation - achieves state-of-the-art 92% on HumanEval (matched only by Claude 3.5 Sonnet). Beats many closed models on MATH problem-solving (71.5%, surpassing Gemini 1.5 Pro, Gemini 1.0 Ultra, GPT-4, Claude 3 Opus). Supports dozens of languages and 80+ coding languages. Features 128K context window for long-context applications. Performs on par with GPT-4o, Claude 3 Opus, and Llama 3 405B while enabling efficient single-node deployment.
Strengths
- State-of-the-art code generation - 92% HumanEval (tied with Claude 3.5 Sonnet)
- Exceptional math performance - 71.5% MATH (beats Gemini 1.5 Pro, GPT-4, Claude 3 Opus)
- Multi-language coding - 76.9% average across Python, C++, Bash, Java, TypeScript, PHP, C#
- 128K context window for long-context applications
- Designed for single-node inference at large throughput (efficient deployment)
- Supports dozens of natural languages and 80+ programming languages
- Performs on par with GPT-4o, Claude 3 Opus, Llama 3 405B
Caveats
- Text-only (no vision or multimodal capabilities)
- 4K max output tokens lower than newer flagship models
- 123B size requires significant resources despite single-node optimization
- Higher cost than Mistral Small ($2 vs $1 input)
- Ecosystem less mature than OpenAI/Anthropic
- Vendor
- Mistral AI
- Context Window
- 128,000 tokens
- Max Output
- 4,096 tokens
- License
- Proprietary
- Release Date
- 2024-07-24
- Modalities
- text
Capabilities
Vision
Audio
Video
Tool Use
Resources
Pricing
$2.00 per 1M input tokens, $6.00 per 1M output tokens