Can Qwen3 Coder 30B A3B Instruct run on the RTX 6000 Ada?
Yes. At 8K context the best fit is Q8_0, using about 31.7 GB of the 45.6 GB usable on the RTX 6000 Ada. Estimated speed ceiling: ~301 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | 11.6 GB Fits | 11.9 GB Fits | 14.2 GB Fits | 23.2 GB Fits | 868 |
| 3-bit (Q3_K_M) • | 14.8 GB Fits | 15.1 GB Fits | 17.4 GB Fits | 26.4 GB Fits | 664 |
| 4-bit (Q4_K_M) • | 18.3 GB Fits | 18.7 GB Fits | 21.0 GB Fits | 30.0 GB Fits | 527 |
| 5-bit (Q5_K_M) • | 21.3 GB Fits | 21.7 GB Fits | 23.9 GB Fits | 32.9 GB Fits | 450 |
| 6-bit (Q6_K) • | 24.4 GB Fits | 24.8 GB Fits | 27.1 GB Fits | 36.1 GB Fits | 389 |
| 8-bit (Q8_0) • | 31.3 GB Fits | 31.7 GB Fits | 33.9 GB Fits | 42.9 GB Tight | 301 |
| 16-bit (F16/BF16) • | 58.0 GB No | 58.3 GB No | 60.6 GB No | 69.6 GB No | — |
• real GGUF file size from unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF. Speed is a bandwidth ceiling (960 GB/s), not a benchmark. Methodology.
More
Qwen3 Coder 30B A3B Instruct on every GPU and Mac · Every model that runs on the RTX 6000 Ada · Other models that fit 48 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (Qwen/Qwen3-Coder-30B-A3B-Instruct), NVIDIA specs. Data as of 2026-11-17.