{"slug": "mistral-large-4-weights-are-out-what-it-takes-to-run-le-chonk", "title": "Mistral Large 4 Weights Are Out: What It Takes to Run Le Chonk", "summary": "Mistral released the weights for Mistral Large 4, a 1.05 trillion-parameter mixture-of-experts model nicknamed \"Le Chonk,\" on October 29, making it the largest open-weight model published by any European or American lab. About 52B parameters activate per forward pass, and the model carries a 1M-token context window, native text-and-image input, 160+ languages, and three reasoning modes, with API pricing of $0.70 per million input tokens and $2.10 per million output tokens. Mistral's announcement reports the model leading FrontierCode 2.0 at 19.7% and Terminal-Bench 4.0 at 41% and tying GLM-5.3 at 48% on the Artificial Analysis Coding Agent Index while using 47% fewer tokens per task, though self-hosting requires at minimum an 8x H200 node at roughly $63/hr for FP8 weights.", "body_md": "Mistral released the weights for its largest model on October 29. At 1.05 trillion parameters, it is the biggest open-weight model any European or American lab has published. The name is **Mistral Large 4**, the nickname is **Le Chonk**, and the pitch is not that it beats closed models — it is that you can download it, fine-tune it, and run it without a vendor who can cut access.\n\n## Architecture: Bigger Numbers, Smarter Routing\n\nLe Chonk is a mixture-of-experts model: 1.05T total parameters but only about 52B activate per forward pass. That architecture is why the per-token compute is tractable at all and why the API pricing ($0.70 per million input tokens, $2.10 output) stays competitive with smaller closed models. The context window is 1M tokens, the model is natively multimodal (text and images), and it covers 160+ languages including every EU official language. Three reasoning modes — Low, Medium, and High — let callers trade latency for depth on a per-request basis.\n\n## Benchmarks: Where It Leads, Where It Does Not\n\nOn the benchmarks Mistral chose to highlight, the results are strong for open-weight systems. [Mistral’s announcement](https://mistral.ai/news/mistral-large-4-oss/) shows Mistral Large 4 leading on FrontierCode 2.0 (19.7%), Terminal-Bench 4.0 (41%), and tying GLM-5.3 on the Artificial Analysis Coding Agent Index at 48% — while using 47% fewer tokens per task. Human evaluators preferred it over GLM-5.3, Kimi K3, and DeepSeek V4.1 Flash in blind pairwise tests. On cybersecurity tasks, it hit a 96.6% malicious-request refusal rate, the highest reported among open models.\n\nThe honest caveat: some critics note the model matches roughly where frontier systems stood in early 2025. The Chinese open-weight labs — DeepSeek, Kimi, GLM — are not standing still. Token efficiency is a meaningful real-world differentiator, but raw benchmark leadership is contested.\n\n## The Hardware Reality\n\nThis is the section most developers actually need. There is no path to running Le Chonk on a single GPU.\n\n- **BF16:** ~2.1 TB — requires three nodes of 8x H100 each\n- **FP8:** ~1.05 TB — one node of 8x H200 (AWS p5en.48xlarge, ~$63/hr)\n- **4-bit AWQ/GPTQ:** ~525 GB — one node of 8x H100, barely fits, minimal KV cache headroom\n- **NVFP4:** ~610 GB — four Blackwell B200 or B300 GPUs; best path on new hardware\n\nThe practical minimum is an 8x H200 node with FP8 weights. The NVFP4 checkpoint is better suited to Blackwell-generation hardware. Per [Qovery’s self-hosting cost guide](https://www.qovery.com/blog/self-host-mistral-large-4-coding-agents-gpu-vllm-cost), 8x H100 at business hours only runs around $11,000 per month; 24/7 is roughly $40,000.\n\n## Spinning It Up With vLLM\n\nOnce the hardware is in place, the vLLM invocation is straightforward:\n\n```\nvllm serve mistralai/Mistral-Large-4.0-1T05-A52B \\\n  --tensor-parallel-size 8 \\\n  --quantization fp8 \\\n  --max-model-len 128000 \\\n  --enable-prefix-caching \\\n  --gpu-memory-utilization 0.92 \\\n  --enable-auto-tool-choice \\\n  --tool-call-parser mistral \\\n  --served-model-name mistral-large-4 \\\n  --port 8000\n```\n\nSGLang is a valid alternative. The model exposes an OpenAI-compatible endpoint, so existing tooling works without changes.\n\n## API vs. Self-Host: The Cost Equation\n\nSelf-hosting only makes financial sense above 20–40% sustained GPU utilization. Most enterprise clusters run around 5% average utilization — meaning you are paying for idle capacity most of the time. The API is cheaper for the majority of organizations.\n\n| Option | Monthly cost | Data control | \n|---|---|---|\n| Mistral API | Variable ($0.70/M in) | None | \n| 8x H100, business hours | ~$11,000 | Full | \n| 8x H100, 24/7 | ~$40,000 | Full | \n\nSpot instances via SkyPilot can cut self-host costs by roughly 70% if interruption tolerance exists. Overnight scale-to-zero via KubeAI or KEDA reduces the business-hours number further.\n\n## The License and the Real Use Case\n\nThe [modified MIT license](https://techcrunch.com/2026/10/29/mistral-releases-weights-for-its-trillion-parameter-ai-model-targeting-cybersecurity/) permits commercial self-hosting, internal tooling, fine-tuning, and product embedding. The one restriction: if you resell model access as a service to third parties — meaning users control inputs or parameters — you need a separate commercial agreement with Mistral.\n\nThat is a reasonable constraint. Mistral VP of Science Pierre Stock framed the value: “Access can’t be cut off by definition because the model is open-weighted.” The model was trained on European infrastructure, covers all EU official languages, and is targeted at organizations where GDPR compliance and data residency make vendor dependency a legal problem, not just a business one.\n\nThe weights are on [Hugging Face](https://huggingface.co/mistralai/Mistral-Large-4-1T-A52B). This is not a model for hobbyists with a single GPU. It is for organizations that already operate multi-GPU clusters and need a frontier-class system they can fully own. If that description fits, the [official weights announcement](https://mistral.ai/news/mistral-large-4-oss/) has the full benchmark breakdown and licensing terms.", "url": "https://wpnews.pro/news/mistral-large-4-weights-are-out-what-it-takes-to-run-le-chonk", "canonical_source": "https://byteiota.com/mistral-large-4-open-weight-deploy/", "published_at": "2026-11-05 20:07:09+00:00", "updated_at": "2026-11-05 20:23:20.119456+00:00", "lang": "en", "topics": ["large-language-models", "artificial-intelligence", "ai-infrastructure", "ai-chips", "ai-tools"], "entities": ["Mistral", "Mistral Large 4", "Le Chonk", "GLM-5.3", "Kimi K3", "DeepSeek V4.1 Flash", "vLLM", "Qovery"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/mistral-large-4-weights-are-out-what-it-takes-to-run-le-chonk", "markdown": "https://wpnews.pro/news/mistral-large-4-weights-are-out-what-it-takes-to-run-le-chonk.md", "text": "https://wpnews.pro/news/mistral-large-4-weights-are-out-what-it-takes-to-run-le-chonk.txt", "jsonld": "https://wpnews.pro/news/mistral-large-4-weights-are-out-what-it-takes-to-run-le-chonk.jsonld"}}