modelgrep

DeepSeek: DeepSeek V3.1 vs Qwen: Qwen3 235B A22B Thinking 2507

Qwen: Qwen3 235B A22B Thinking 2507 wins on more metrics (5 of 9), but the right pick depends on what you optimize for — see the breakdown below.

MetricDeepSeek: DeepSeek V3.1Qwen: Qwen3 235B A22B Thinking 2507
Intelligence Index28.129.5
Coding Index28.423.2
GPQA Diamond74%79%
Design Arena Elo11671097
Speed (tokens/sec)9672
Latency399ms508ms
Input price /M$0.210$0.100
Output price /M$0.790$0.100
Context window164K262K
CapabilitiesReasoningToolsJSONReasoningToolsJSON