modelgrep

Anthropic: Claude Sonnet 4.5 vs Qwen: Qwen3.6 27B

Qwen: Qwen3.6 27B wins on more metrics (5 of 8), but the right pick depends on what you optimize for — see the breakdown below.

MetricAnthropic: Claude Sonnet 4.5Qwen: Qwen3.6 27B
Intelligence Index37.137.1
Coding Index33.526.6
GPQA Diamond73%83%
Design Arena Elo1242
Speed (tokens/sec)4376
Latency1.2s532ms
Input price /M$3.00$0.288
Output price /M$15.00$3.17
Context window1M262K
CapabilitiesReasoningToolsJSONVisionReasoningToolsJSONVision