Alibaba · 80B (3B active) · Mixture of Experts
Ultra-efficient agentic coding MoE optimized for tool-calling coding agents
0 downloads
0 likes
2026-02 256K context
Use Cases
code
Mixture of Experts
Total experts: 512
Active experts: 10
Active params: 3.0B
| Quant | Bits | VRAM | Quality | Status |
|---|---|---|---|---|
| Q2_K | 2 | 26.1 GB | low | — |
| Q3_K_M | 3 | 36.4 GB | moderate | — |
| Q4_K_M | 4 | 41.5 GB | good | — |
| Q5_K_M | 5 | 51.7 GB | good | — |
| Q6_K | 6 | 62 GB | excellent | — |
| Q8_0 | 8 | 82.5 GB | excellent | — |
| F16 | 16 | 164.4 GB | lossless | — |
GPUs that run Qwen 3 Coder Next 80B-A3B
Qwen 3 Coder Next 80B-A3B needs about 41.5 GB of VRAM at Q4_K_M, so it runs comfortably on a GPU with 80 GB of VRAM or more (10 of the GPUs we track). Common cards, smallest first:
| VRAM | GPUs |
|---|---|
| 80 GB | A100 , H100 |
| 96 GB | RTX PRO 6000 , GH200 , Ryzen AI MAX+ 395 |
| 128 GB | DGX Spark |
| 141 GB | H200 |
About this model
Ultra-efficient agentic coding MoE optimized for tool-calling coding agents