L40S
2 records across 1 models (quants: Ternary PQ2_0, Ternary PTQ1_0).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| L40S | Ternary-Bonsai-2-27B | Ternary PTQ1_0 | llama.cpp (CUDA) | 81.8 | 1543 | – | – | tg128 | 2026-09-16 | community | – | source |
| L40S | Ternary-Bonsai-2-27B | Ternary PQ2_0 | llama.cpp (CUDA) | 74.4 | 2868 | – | – | tg128 | 2026-09-16 | community | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085