{"slug": "mistral-large-4-is-open-weight-read-the-license-and-do-the-memory-math-before", "title": "Mistral Large 4 Is Open Weight. Read the License and Do the Memory Math Before You Self-Host", "summary": "Mistral released the open weights of Mistral Large 4, a roughly 1-trillion-parameter natively multimodal Mixture-of-Experts model with a 1M-token context, on Hugging Face on October 29, 2026. The release ships under the modified-MIT Mistral Ecosystem Open-Weight License, which requires a separate commercial license for anyone operating a Model-as-a-Service business, and self-hosting demands roughly 1 TB of weights at FP8 or about 0.5 TB at NVFP4 because all 128 experts must stay resident even though only 52B parameters are active per token. Mistral reports 41% on Terminal-Bench 4.0 and 48% on the Artificial Analysis Coding Agent Index, though those figures remain vendor claims pending independent evaluation.", "body_md": "On October 29, 2026, Mistral released the weights of **Mistral Large 4** (nicknamed \"Le Chonk\") on Hugging Face. It's a roughly 1 trillion parameter, natively multimodal Mixture-of-Experts model with a 1M-token context, and Mistral calls it the best open-weight model for coding.\n\n\"Open weight\" headlines make self-hosting sound like a download away. Two details decide whether it actually is for your team: **what the license lets you do**, and **how much hardware \"52B active\" really needs**. This post walks through both, plus how I'd decide between the API and running it yourself.\n\n*Benchmark figures below are Mistral's own reported numbers unless noted. Treat them as claims until independent evals land.*\n\n`mistralai/Mistral-Large-4-1T-A52B` on Hugging Face, in two official variants: Mistral's reported results include 19.7% on Cognition's private FrontierCode 2.0, 41% on Terminal-Bench 4.0 (up 13 points from the preview while using 30% fewer output tokens), 48% on the Artificial Analysis Coding Agent Index (roughly tied with GLM-5.3 while using 47% fewer tokens per task), 84.5% on BrowseComp, and 97.1% attack resistance on the B3 Agent Security benchmark.\n\nThe license is the **Mistral Ecosystem Open-Weight License**, a modified MIT license. It grants broad rights to use, copy, modify, fine-tune, distribute and sell the model and derivatives. Then section 2 adds the condition that matters:\n\nIf the Licensee or any of its affiliates operates a Model as a Service business, the Licensee must enter into a separate commercial license with Mistral before using the Model or its derivative works for any commercial purpose.\n\n\"Model as a Service\" means giving a third party access to the model for inference or post-training (an API, a hosted endpoint) **in a way that lets them exercise meaningful control over the inputs, parameters, or training data**.\n\nThe license carves out:\n\nIn practice, a quick way to read it:\n\n| What you're building | Likely bucket | \n|---|---|\n| Internal coding assistant or RAG over company docs | Internal use, no extra license | \n| SaaS feature: \"summarize this ticket\" button | Embedded in a specific feature | \n| Public API where customers send their own prompts and params | Model as a Service, needs a commercial license | \n| Fine-tuning service for customers' data | Model as a Service (post-training) | \n\nNote the condition is about whether **you or an affiliate operate** a MaaS business, and there's no revenue threshold. If your company runs any hosted-model offering, talk to legal before you ship. I'm not a lawyer; read the full license text linked below.\n\n\"52B active\" sounds like a mid-size model. It isn't, from a memory point of view.\n\nIn an MoE, the router picks 4 of 128 experts **per token**. Different tokens pick different experts, so every expert has to be loaded and ready. That means:\n\nRough weight-only arithmetic (my estimate, not Mistral's figure):\n\n```\nFP8   : ~1 byte per param   x ~1T params  = ~1 TB of weights\nNVFP4 : ~0.5 byte per param x ~1T params  = ~0.5 TB of weights (+ scale overhead)\n```\n\nOn top of that you need KV cache, and a 1M-token context makes KV cache a serious line item if you actually use long prompts. This is a multi-GPU, multi-node deployment, which is why the Hugging Face example sets both tensor parallel (`tp_size`) and expert parallel (` ep_size`) sizes. GGUF quants make local experiments possible, but heavy quantization can cost quality, so re-run your evals on whatever quant you deploy.\n\nOpen weights give you options, not a free lunch. Read the license for your exact product shape, size hardware by total parameters, and prove the model wins on your own tasks through the API before you buy a rack of GPUs for it.\n\nWould you self-host a 1T MoE model, or is the API good enough for your use case?", "url": "https://wpnews.pro/news/mistral-large-4-is-open-weight-read-the-license-and-do-the-memory-math-before", "canonical_source": "https://dev.to/m_t_ramkrushna/mistral-large-4-is-open-weight-read-the-license-and-do-the-memory-math-before-you-self-host-4dij", "published_at": "2026-11-03 04:07:13+00:00", "updated_at": "2026-11-03 04:22:13.081929+00:00", "lang": "en", "topics": ["large-language-models", "artificial-intelligence", "ai-infrastructure", "ai-tools"], "entities": ["Mistral", "Hugging Face", "Mistral Large 4", "Cognition", "Terminal-Bench", "Artificial Analysis", "GLM-5.3"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/mistral-large-4-is-open-weight-read-the-license-and-do-the-memory-math-before", "markdown": "https://wpnews.pro/news/mistral-large-4-is-open-weight-read-the-license-and-do-the-memory-math-before.md", "text": "https://wpnews.pro/news/mistral-large-4-is-open-weight-read-the-license-and-do-the-memory-math-before.txt", "jsonld": "https://wpnews.pro/news/mistral-large-4-is-open-weight-read-the-license-and-do-the-memory-math-before.jsonld"}}