mistralai/Mistral-Large-4-1T-A52B
AI model · https://huggingface.co/mistralai/Mistral-Large-4-1T-A52B
First seen by Oriel on .
AI summary
This model card presents Mistral Large 4, a 1-trillion-parameter multimodal Mixture-of-Experts model for reasoning, coding, vision, multilingual, and agentic workloads.
- It uses 128 routed experts, four active per token, one shared expert, approximately 52 billion active parameters, and a 1M-token context window.
- The model accepts text and variable-sized images, supports configurable low, medium, and high reasoning effort, function calls, JSON schemas, and 160+ languages.
- On reported benchmarks, it tied GLM-5.3 at 48% on the Artificial Analysis Coding Agent Index while using 47% fewer tokens per task.
- Its embedded DSpark draft head proposes three tokens and delivered up to 1.9x higher throughput without changing target-model outputs, depending on hardware and workload.
- Deployment is supported through vLLM, SGLang, Transformers, and llama.cpp; a quantized NVFP4 checkpoint reduces memory usage and targets Blackwell GPUs.
#mixture of experts #multimodal ai #llm inference #agentic ai
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Likes
Total likes reported by the source.
- Hugging Face Models: 1,646, latest observation ; +561 over observed period since
Downloads (30d)
Rolling 30-day downloads. A change in this value is not new downloads during the observed period.
- Hugging Face Models: 12,541, latest observation ; +12,263 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Models | mistralai/Mistral-Large-4-1T-A52B | https://huggingface.co/mistralai/Mistral-Large-4-1T-A52B Read discussion |