Benchmark result
qwen3.8-flash-next-unsloth-ud-q4_k_xl on AMD Radeon RX 7900 XTX — 65.2 tok/s
Measured with Strata Strata 0.1.59 on November 15, 2026.
Support llm-bench.io
llm-bench.io is free. If it helped you, I'd appreciate your support to cover the infrastructure costs.
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; the frame only adds a line that tells this page how tall it is.
How this run compares
2nd fastest of 11 runs of this model on AMD Radeon RX 7900 XTX · median 57.4 tok/s.
sort
- 165.9tok/s85.4Strata · Q4_K
- 265.2tok/s84.9Strata · Q4_Kthis run
- 364.2tok/s77.5Strata · Q4_K
- 463.8tok/s80.7Strata · Q4_K
- 557.8tok/s82.8Strata · Q4_K
- 657.4tok/s89.1Strata · Q4_K
- 757.0tok/s86.9Strata · Q4_K
- 857.0tok/s78.6Strata · Q4_K
- 955.1tok/s73.6Strata · Q4_K
- 1055.1tok/s85.8Strata · Q4_K
- 1140.8tok/s83.0lama.cpp · Q4_K
qwen3.8-flash-next-unsloth-ud-q4_k_xl on other hardware
Reproduce this run
toolStrata Strata 0.1.59
modelqwen3.8-flash-next-unsloth-ud-q4_k_xl (Q4_K)
context327,680
thinkingon
effortmedium
samplingtemperature 1 · top_p 0.95 · top_k 20
clientv0.4.99+102
Set those in Strata, then:
llm-benchmark benchmark --model "qwen3.8-flash-next-unsloth-ud-q4_k_xl" --tool "Strata"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.