Can North Mini Code 1.0 run on the A100 40GB?
Yes. At 8K context the best fit is Q8_0, using about 31.6 GB of the 38.0 GB usable on the A100 40GB. Estimated speed ceiling: ~442 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) | 11.9 GB Fits | 12.0 GB Fits | 12.6 GB Fits | 15.1 GB Fits | 1255 |
| 3-bit (Q3_K_M) • | 14.5 GB Fits | 14.6 GB Fits | 15.2 GB Fits | 17.7 GB Fits | 1009 |
| 4-bit (Q4_K_M) • | 19.2 GB Fits | 19.3 GB Fits | 19.9 GB Fits | 22.3 GB Fits | 747 |
| 5-bit (Q5_K_M) • | 22.7 GB Fits | 22.8 GB Fits | 23.4 GB Fits | 25.8 GB Fits | 625 |
| 6-bit (Q6_K) • | 25.0 GB Fits | 25.1 GB Fits | 25.8 GB Fits | 28.2 GB Fits | 562 |
| 8-bit (Q8_0) • | 31.5 GB Fits | 31.6 GB Fits | 32.2 GB Fits | 34.6 GB Tight | 442 |
| 16-bit (F16/BF16) • | 58.1 GB No | 58.2 GB No | 58.8 GB No | 61.2 GB No | — |
• real GGUF file size from unsloth/North-Mini-Code-1.0-GGUF. Speed is a bandwidth ceiling (1555 GB/s), not a benchmark. Methodology.
More
North Mini Code 1.0 on every GPU and Mac · Every model that runs on the A100 40GB · Other models that fit 48 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (CohereLabs/North-Mini-Code-1.0), NVIDIA specs. Data as of 2026-11-17.