modelgrep

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

Qwen: Qwen3 235B A22B Thinking 2507 wins on more metrics (7 of 9), but the right pick depends on what you optimize for — see the breakdown below.

MetricDeepSeek: DeepSeek V3.1 TerminusQwen: Qwen3 235B A22B Thinking 2507
Intelligence Index28.529.5
Coding Index31.923.2
GPQA Diamond75%79%
Design Arena Elo12381097
Speed (tokens/sec)2672
Latency933ms508ms
Input price /M$0.270$0.100
Output price /M$0.950$0.100
Context window164K262K
CapabilitiesReasoningToolsJSONReasoningToolsJSON