Benchmark result

ukisai/Swift-1.5-Qwen3.8-27B-GGUF:Swift-1.5-Qwen3.8-27B-IQ4_XS on AMD Radeon RX 7900 XT/7900 XTX/7900 GRE/7900M +1 more — 71.1 tok/s

Measured with llama.cpp b11195-d834d44e6 on September 27, 2026.

What the model built

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 AMD Radeon RX 7900 XT/7900 XTX/7900 GRE/7900M +1 more · Multi-GPU · median 71.5 tok/s.

sort

171.9tok/s87.8llama.cpp
271.1tok/s79.2llama.cppthis run

Compare the nearest 5 →

Reproduce this run

toolllama.cpp b11195-d834d44e6

modelukisai/Swift-1.5-Qwen3.8-27B-GGUF:Swift-1.5-Qwen3.8-27B-IQ4_XS

context65,536

thinkingon

effortxhigh (model default)

samplingtemperature 1 · top_p 0.949999988079071 · top_k 20 · min_p 0.05000000074505806 · presence_penalty 0 · repeat_penalty 1

clientv0.4.80+98

Set those in llama.cpp, then:

llm-benchmark benchmark --model "ukisai/Swift-1.5-Qwen3.8-27B-GGUF:Swift-1.5-Qwen3.8-27B-IQ4_XS" --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.