Can Trinity Mini run on the L40S?
Yes. At 8K context the best fit is Q8_0, using about 26.9 GB of the 45.6 GB usable on the L40S. Estimated speed ceiling: ~413 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) • | 9.7 GB Fits | 9.8 GB Fits | 10.2 GB Fits | 11.7 GB Fits | 1219 |
| 3-bit (Q3_K_M) • | 12.2 GB Fits | 12.3 GB Fits | 12.6 GB Fits | 14.1 GB Fits | 949 |
| 4-bit (Q4_K_M) • | 15.8 GB Fits | 15.8 GB Fits | 16.2 GB Fits | 17.7 GB Fits | 721 |
| 5-bit (Q5_K_M) • | 18.3 GB Fits | 18.4 GB Fits | 18.7 GB Fits | 20.2 GB Fits | 616 |
| 6-bit (Q6_K) • | 21.0 GB Fits | 21.0 GB Fits | 21.4 GB Fits | 22.9 GB Fits | 534 |
| 8-bit (Q8_0) • | 26.8 GB Fits | 26.9 GB Fits | 27.3 GB Fits | 28.8 GB Fits | 413 |
| 16-bit (F16/BF16) | 49.6 GB No | 49.7 GB No | 50.0 GB No | 51.5 GB No | — |
• real GGUF file size from arcee-ai/Trinity-Mini-GGUF. Speed is a bandwidth ceiling (864 GB/s), not a benchmark. Methodology.
More
Trinity Mini 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 (arcee-ai/Trinity-Mini), NVIDIA specs. Data as of 2026-11-17.