Can Maxwell V1 Flash run on the A100 40GB?
Yes. At 8K context the best fit is Q6_K, using about 29.2 GB of the 38.0 GB usable on the A100 40GB. Estimated speed ceiling: ~835 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) | 13.8 GB Fits | 14.0 GB Fits | 15.0 GB Fits | 19.3 GB Fits | 1826 |
| 3-bit (Q3_K_M) | 17.7 GB Fits | 17.9 GB Fits | 18.9 GB Fits | 23.2 GB Fits | 1401 |
| 4-bit (Q4_K_M) | 21.9 GB Fits | 22.1 GB Fits | 23.1 GB Fits | 27.4 GB Fits | 1120 |
| 5-bit (Q5_K_M) | 25.4 GB Fits | 25.6 GB Fits | 26.6 GB Fits | 30.8 GB Fits | 961 |
| 6-bit (Q6_K) | 29.1 GB Fits | 29.2 GB Fits | 30.3 GB Fits | 34.5 GB Tight | 835 |
| 8-bit (Q8_0) • | 36.7 GB Tight | 36.9 GB Tight | 38.0 GB Tight | 42.2 GB No | 657 |
| 16-bit (F16/BF16) • | 68.0 GB No | 68.2 GB No | 69.3 GB No | 73.5 GB No | — |
• real GGUF file size from AtomicChat/Maxwell-V1-Flash-GGUF. Speed is a bandwidth ceiling (1555 GB/s), not a benchmark. Methodology.
More
Maxwell V1 Flash 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 (maxwell-model/Maxwell-V1-Flash), NVIDIA specs. Data as of 2026-11-17.