Check my GPU

LLMs ·

Qwen3.8 Max (2.4T-A95B)

Parameters
2,446B (95.0B active)
Architecture
Mixture of experts
Max context
256K tokens
License
qwen3.8-max
Released
2026-08

How to run it

Quantized files · GGUF

Works in LM Studio, llama.cpp, Jan and Ollama. No sign-in needed.

huggingface.co/unsloth/Qwen3.8-2.4T-A95B-GGUF (opens in a new tab)

ollama run hf.co/unsloth/Qwen3.8-2.4T-A95B-GGUF

Original weights · Hugging Face

Full-precision files from Alibaba, for fine-tuning, vLLM, Transformers.

huggingface.co/Qwen/Qwen3.8-2.4T-A95B (opens in a new tab)

No sign-in needed.

Official repo · GitHub

Reference code, inference scripts and docs from Alibaba. The weights themselves download from Hugging Face.

github.com/QwenLM/Qwen3.8 (opens in a new tab)

git clone https://github.com/QwenLM/Qwen3.8

Qwen’s current repo: Qwen3.5 and Qwen3.6 redirect here.

Links checked 7 days ago against Hugging Face, Ollama and GitHub.

Memory needed

Weights plus KV cache and runtime overhead. Leave a little extra free for the operating system and other apps.

KV cache: hybrid: only the full-attention layers (1 in 4) keep a KV cache; the rest use linear attention.

Quantization Quality Weights Total at 8K context Total at 32K context Download
Q2_K Low 978 GB 980 GB 982 GB GGUF (IQ2_XXS) · folder (opens in a new tab)
Q3_K_M Low 1,199 GB 1,200 GB 1,202 GB GGUF (IQ3_XXS) · folder (opens in a new tab)
Q4_K_M Good 1,492 GB 1,494 GB 1,496 GB GGUF files (opens in a new tab)
Q5_K_M High 1,737 GB 1,738 GB 1,740 GB GGUF files (opens in a new tab)
Q6_K High 2,006 GB 2,007 GB 2,010 GB GGUF files (opens in a new tab)
Q8_0 Very high 2,593 GB 2,594 GB 2,597 GB GGUF · folder (opens in a new tab)
F16 Full 4,892 GB 4,894 GB 4,896 GB GGUF · folder (opens in a new tab) Original weights (opens in a new tab)

Best hardware for Qwen3.8 Max (2.4T-A95B)

None of the popular GPUs or Macs in our catalog can run this model. It needs datacenter hardware or several GPUs.

Check it on your own hardware

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.