How much VRAM does Qwen3 Coder Next 80B need?
Qwen3 Coder Next 80B needs 45.0 GB for its weights at 4-bit, 46.0 GB with an 8K context and 46.6 GB with 32K. At 8-bit, 79.9 GB.
- Parameters
- 79.7B
- Active per token
- 3B
- Longest context
- 256K
- Released
- Jan 2026
- License
- Apache 2.0
- Strongest at
- Coding, Agents
Memory at each precision
| Precision | Weights | With 8K | With 32K | With 128K |
|---|---|---|---|---|
| 16-bit | 148 GB | 149 GB | 150 GB | 152 GB |
| 8-bit | 78.8 GB | 79.9 GB | 80.5 GB | 82.7 GB |
| 6-bit | 60.8 GB | 61.9 GB | 62.5 GB | 64.7 GB |
| 5-bit | 52.8 GB | 53.8 GB | 54.4 GB | 56.6 GB |
| 4-bit | 45.0 GB | 46.0 GB | 46.6 GB | 48.9 GB |
| 3-bit | 36.3 GB | 37.3 GB | 37.9 GB | 40.1 GB |
| 2-bit | 27.8 GB | 28.9 GB | 29.4 GB | 31.7 GB |
| 1.6-bit | 15.3 GB | 16.4 GB | 16.9 GB | 19.2 GB |
One conversation, 16-bit context cache, runtime overhead included.
What runs it
| Machine | Best fit | Context | Speed | First reply |
|---|---|---|---|---|
| RTX 3060 12 GB | Too big | |||
| RTX 4060 Ti 16 GB | Too big | |||
| RTX 3090 | Too big | |||
| RTX 4090 | Too big | |||
| RTX 5090 | Too big | |||
| 2× RTX 3090 | 3-bit | 256K | 191 tok/s | 0.3 s |
| RTX PRO 6000 Blackwell | 8-bit | 256K | 182 tok/s | 0.3 s |
| M4 Max 128 GB | 8-bit | 256K | 63 tok/s | 2.9 s |
| M5 Max 128 GB | 8-bit | 256K | 67 tok/s | 0.9 s |
| Ryzen AI Max+ 395 128 GB | 8-bit | 256K | 39 tok/s | 2.2 s |
| DGX Spark 128 GB | 8-bit | 256K | 50 tok/s | 1.8 s |
| M3 Ultra 256 GB | 8-bit | 256K | 78 tok/s | 1.7 s |
| M5 Ultra 512 GB | 8-bit | 256K | 92 tok/s | 0.5 s |
| H100 SXM | 6-bit | 256K | 252 tok/s | 0.1 s |
| DGX Station GB300 | 8-bit | 256K | 283 tok/s | 0.1 s |
| 8× H200 NVL | 8-bit | 256K | 260 tok/s | 0.1 s |
Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.
How to run it
- Ollama
ollama run qwen3-coder-next:latest- llama.cpp
llama-server -hf unsloth/Qwen3-Coder-Next-GGUF- vLLM
vllm serve Qwen/Qwen3-Coder-Next- MLX (Mac)
mlx_lm.chat --model mlx-community/Qwen3-Coder-Next-4bit
Questions
How much VRAM does Qwen3 Coder Next 80B need?
Qwen3 Coder Next 80B needs 45.0 GB for its weights at 4-bit, 46.0 GB with an 8K context and 46.6 GB with 32K. At 8-bit, 79.9 GB.
Can Qwen3 Coder Next 80B run on an RTX 4090?
Not on one card. The smallest machine here that runs it is the 2× RTX 3090.