Can HunyuanOCR run on the L40S?
Yes. At 8K context the best fit is F16, using about 3.5 GB of the 45.6 GB usable on the L40S. Estimated speed ceiling: ~386 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) | 1.4 GB Fits | 1.8 GB Fits | 4.0 GB Fits | 13.0 GB Fits | 2057 |
| 3-bit (Q3_K_M) | 1.5 GB Fits | 1.9 GB Fits | 4.1 GB Fits | 13.1 GB Fits | 1578 |
| 4-bit (Q4_K_M) | 1.6 GB Fits | 2.0 GB Fits | 4.3 GB Fits | 13.3 GB Fits | 1262 |
| 5-bit (Q5_K_M) | 1.7 GB Fits | 2.1 GB Fits | 4.4 GB Fits | 13.4 GB Fits | 1083 |
| 6-bit (Q6_K) | 1.9 GB Fits | 2.2 GB Fits | 4.5 GB Fits | 13.5 GB Fits | 941 |
| 8-bit (Q8_0) | 2.1 GB Fits | 2.5 GB Fits | 4.7 GB Fits | 13.7 GB Fits | 726 |
| 16-bit (F16/BF16) | 3.1 GB Fits | 3.5 GB Fits | 5.7 GB Fits | 14.7 GB Fits | 386 |
• real GGUF file size. Speed is a bandwidth ceiling (864 GB/s), not a benchmark. Methodology.
More
HunyuanOCR on every GPU and Mac · Every model that runs on the L40S · Other models that fit 48 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (tencent/HunyuanOCR), NVIDIA specs. Data as of 2026-11-17.