Benchmark result
Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_S-mtp on 2× NVIDIA GeForce RTX 5060 — 37.5 tok/s
Measured with llama.cpp b11130-183d2a04c on September 25, 2026.
What the model built
Open full screen →The coding scenario asks for a playable game in a single HTML file. This is exactly what the model returned, unedited.
How this run compares
2nd fastest of 2 runs of this model on 2× NVIDIA GeForce RTX 5060 · Multi-GPU · median 37.9 tok/s.
sort
- 138.3tok/s87.0llama.cpp
- 237.5tok/s77.8llama.cppthis run
Reproduce this run
toolllama.cpp b11130-183d2a04c
modelSwift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_S-mtp
context98,304
thinkingon
samplingtemperature 0.699999988079071 · top_p 0.949999988079071 · top_k 20 · min_p 0.05999999865889549 · presence_penalty 0 · repeat_penalty 1.0499999523162842
clientv0.4.80+98
Set those in llama.cpp, then:
llm-benchmark benchmark --model "Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_S-mtp" --tool "llama.cpp"The client prompts for context length and thinking mode, and for the KV cache dtype on Unsloth Studio. Sampling is left at the model default — the values above are what the tool reported using, not overrides the client sent.