Can Hermes 4.3 36B run on the A100 80GB?
Yes. At 8K context the best fit is Q8_0, using about 38.5 GB of the 76.0 GB usable on the A100 80GB. Estimated speed ceiling: ~53 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) • | 14.4 GB Fits | 15.4 GB Fits | 21.4 GB Fits | 45.4 GB Fits | 150 |
| 3-bit (Q3_K_M) • | 18.1 GB Fits | 19.1 GB Fits | 25.1 GB Fits | 49.1 GB Fits | 116 |
| 4-bit (Q4_K_M) • | 22.0 GB Fits | 23.0 GB Fits | 29.0 GB Fits | 53.0 GB Fits | 94 |
| 5-bit (Q5_K_M) • | 25.5 GB Fits | 26.5 GB Fits | 32.5 GB Fits | 56.5 GB Fits | 80 |
| 6-bit (Q6_K) • | 29.3 GB Fits | 30.3 GB Fits | 36.3 GB Fits | 60.3 GB Fits | 69 |
| 8-bit (Q8_0) • | 37.5 GB Fits | 38.5 GB Fits | 44.5 GB Fits | 68.5 GB Tight | 53 |
| 16-bit (F16/BF16) • | 69.0 GB Tight | 70.0 GB Tight | 76.0 GB No | 100 GB No | 28 |
• real GGUF file size from bartowski/NousResearch_Hermes-4.3-36B-GGUF. Speed is a bandwidth ceiling (2039 GB/s), not a benchmark. Methodology.
More
Hermes 4.3 36B on every GPU and Mac · Every model that runs on the A100 80GB · Other models that fit 64 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (NousResearch/Hermes-4.3-36B), NVIDIA specs. Data as of 2026-11-17.