Benchmark result
openai/gpt-oss-20b on AMD Radeon RX 7900 XTX — 178.1 tok/s
Measured with LM Studio on October 23, 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; the frame only adds a line that tells this page how tall it is.
How this run compares
1st fastest of 21 runs of this model on AMD Radeon RX 7900 XTX · median 70.6 tok/s.
sort
- 1178.1tok/s77.9LM Studio · MXFP4this run
- 2111.1tok/s63.8LM Studio
- 3111.0tok/s74.6LM Studio
- 4110.8tok/s75.8LM Studio
- 5110.0tok/s64.6LM Studio
- 6109.6tok/s64.2LM Studio
- 7108.3tok/s62.5LM Studio
- 8108.3tok/s63.4LM Studio
- 9106.7tok/s67.8LM Studio
- 10106.2tok/s66.2LM Studio
- 1170.6tok/s72.6Ollama · MXFP4
- 1269.1tok/s75.4Ollama · MXFP4
- 1367.8tok/s75.9Ollama · MXFP4
- 1466.7tok/s74.5Ollama · MXFP4
- 1564.9tok/s71.6Ollama · MXFP4
- 1663.5tok/s77.9Ollama · MXFP4
- 1762.1tok/s73.4Ollama · MXFP4
- 1860.5tok/s73.7Ollama · MXFP4
- 1957.0tok/s75.0Ollama · MXFP4
- 2051.3tok/s73.8Ollama · MXFP4
- 2148.6tok/s64.5Ollama · MXFP4
Reproduce this run
toolLM Studio
modelopenai/gpt-oss-20b (MXFP4)
context65,536
thinkingon
efforthigh
clientv0.4.97+102
Set those in LM Studio, then:
llm-benchmark benchmark --model "openai/gpt-oss-20b" --tool "LM Studio"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.