modelgrep

Anthropic: Claude Sonnet 4 vs OpenAI: gpt-oss-120b

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.

MetricAnthropic: Claude Sonnet 4OpenAI: gpt-oss-120b
Intelligence Index33.033.3
Coding Index30.628.6
GPQA Diamond68%78%
Design Arena Elo12201062
Speed (tokens/sec)50554
Latency720ms161ms
Input price /M$3.00$0.039
Output price /M$15.00$0.180
Context window1M131K
CapabilitiesReasoningToolsVisionReasoningToolsJSON