modelgrep

OpenAI: gpt-oss-120b vs Qwen: Qwen3 235B A22B Instruct 2507

OpenAI: gpt-oss-120b wins on more metrics (6 of 9), but the right pick depends on what you optimize for — see the breakdown below.

MetricOpenAI: gpt-oss-120bQwen: Qwen3 235B A22B Instruct 2507
Intelligence Index33.325.0
Coding Index28.622.1
GPQA Diamond78%75%
Design Arena Elo10621103
Speed (tokens/sec)69986
Latency149ms287ms
Input price /M$0.039$0.090
Output price /M$0.180$0.100
Context window131K262K
CapabilitiesReasoningToolsJSONToolsJSON