readygpu

Qwen3 Next 80B-A3B

81.3B total, 3B active · 256k context · apache-2.0 · released 2025-09-09

chatcodingagentic

What to run it on in the UK

Fully in memory at 128k context, priced at retailers today.

Something we price holds it fully in memory, but none reaches 15 tok/s, our bar for recommending it. The table below has the speeds.

All hardware that runs it

Everything we price, best value first. Two cards split the model by layers.

HardwareQuantFitsDecodePricetok/s per 1,000
Mac Studio M4 Max 128GBQ8_0fits33 tok/s––
Framework Desktop Ryzen AI Max+ 395 128GBQ8_0fits13 tok/s––
Mac Studio M3 Ultra 512GBQ8_0fits50 tok/s––
NVIDIA DGX Spark 128GBQ8_0fits16 tok/s––
GMKtec EVO-X2 Ryzen AI Max+ 395 128GBQ8_0fits13 tok/s––
Mac Studio M3 Ultra 256GBQ8_0fits50 tok/s––
Mac Studio M3 Ultra 96GBQ6_Kfits56 tok/s––
NVIDIA RTX PRO 6000 BlackwellQ8_0fits104 tok/s––
ASUS Ascent GX10 128GBQ8_0fits16 tok/s––
NVIDIA GeForce RTX 5090Q4_K_Moffload16 tok/s––
AMD Radeon AI PRO R9700Q4_K_Moffload13 tok/s––

29 other cards and machines we price cannot hold it at this context.

Memory needed

Weights, KV cache and runtime buffers at each context length (KV cache at fp16).

QuantFile@8k@32k@64k@128k
Q4_K_Mgguf45.2 GB46.2 GB46.9 GB47.8 GB49.7 GB
mlx-4bitmlx41.8 GB42.8 GB43.5 GB44.4 GB46.3 GB
Q5_K_Mgguf52.9 GB53.9 GB54.6 GB55.6 GB57.4 GB
Q6_Kgguf61.0 GB62.0 GB62.7 GB63.7 GB65.6 GB
mlx-8bitmlx78.8 GB79.8 GB80.5 GB81.5 GB83.4 GB
Q8_0gguf79.0 GB80.0 GB80.7 GB81.6 GB83.5 GB

Run it

Copy a command for the runtime you use.

  • Q4_K_M
    • Ollamaollama run hf.co/unsloth/Qwen3-Next-80B-A3B-Instruct-GGUF:Q4_K_M
    • LM Studiolms get unsloth/Qwen3-Next-80B-A3B-Instruct-GGUF@q4_k_m
    • llama.cppllama-server -hf unsloth/Qwen3-Next-80B-A3B-Instruct-GGUF:Q4_K_M
  • mlx-4bit
    • LM Studiolms get mlx-community/Qwen3-Next-80B-A3B-Instruct-4bit
    • MLXmlx_lm.server --model mlx-community/Qwen3-Next-80B-A3B-Instruct-4bit
  • Q5_K_M
    • Ollamaollama run hf.co/unsloth/Qwen3-Next-80B-A3B-Instruct-GGUF:Q5_K_M
    • LM Studiolms get unsloth/Qwen3-Next-80B-A3B-Instruct-GGUF@q5_k_m
    • llama.cppllama-server -hf unsloth/Qwen3-Next-80B-A3B-Instruct-GGUF:Q5_K_M
  • Q6_K
    • Ollamaollama run hf.co/unsloth/Qwen3-Next-80B-A3B-Instruct-GGUF:Q6_K
    • LM Studiolms get unsloth/Qwen3-Next-80B-A3B-Instruct-GGUF@q6_k
    • llama.cppllama-server -hf unsloth/Qwen3-Next-80B-A3B-Instruct-GGUF:Q6_K
  • mlx-8bit
    • LM Studiolms get mlx-community/Qwen3-Next-80B-A3B-Instruct-8bit
    • MLXmlx_lm.server --model mlx-community/Qwen3-Next-80B-A3B-Instruct-8bit
  • Q8_0
    • Ollamaollama run hf.co/unsloth/Qwen3-Next-80B-A3B-Instruct-GGUF:Q8_0
    • LM Studiolms get unsloth/Qwen3-Next-80B-A3B-Instruct-GGUF@q8_0
    • llama.cppllama-server -hf unsloth/Qwen3-Next-80B-A3B-Instruct-GGUF:Q8_0