Model · Alibaba Qwen

Qwen3-Coder Next

79.7B MoE · Q4_K_M · 45.1 GB · open weights · curated for coding, agents · repo created 2026-01-30

Too much for a 24 GB card: 48.1 GB at 8K context against 21.1 GB usable. Runs easily on NVIDIA DGX Spark.

What the model needs: weights, context cache and runtime overhead at four reference contexts. The total is the model's own; what differs per machine is the usable memory it has to fit into — VRAM on a graphics card, unified memory on a Mac.

Cache shape from the model's config: 48 of 48 layers cache the full context, each storing K and V for 2 KV heads × 256 dimensions per token at f16. Multiply those four numbers by two bytes and by the context length and you have the cache column.

All memory figures are estimates: measured quantized file size + computed context memory + runtime overhead, with a 12% safety margin on your hardware. Real usage varies with runtime version and settings.

Every tracked GPU and Mac at 8K and 32K context with a full-precision (f16) cache — the same engine as the finder and the GPU pages.

Runs on 3 of 32 tracked devices at 8K context, easily or with ordinary headroom.

GGUF is the quantized single-file format local runtimes load (Ollama, LM Studio, llama.cpp); our memory figures are measured from the linked GGUF file. The original repo holds full-precision safetensors — a much larger download meant for GPUs with far more memory.

More from Alibaba Qwen

• Qwen3.8 Flash-Nextabout the same ◆
• Qwen3.8 27Babout the same ◆
• Qwen3.8 2.4T-A95Bmore popular ▲
• Qwen3 14Babout the same ◆
• Qwen3 0.6Bfading ▼
• Qwen3-Coder 30B-A3Babout the same ◆

Every tracked model →