modelgrep

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

DDeepSeek: DeepSeek V3.1 Terminusdeepseek/deepseek-v3.1-terminusVSQQwen: Qwen3 VL 235B A22B Thinkingqwen/qwen3-vl-235b-a22b-thinking

DeepSeek: DeepSeek V3.1 Terminus wins on more metrics (6 of 9), but the right pick depends on what you optimize for — see the breakdown below.

MetricDeepSeek: DeepSeek V3.1 TerminusQwen: Qwen3 VL 235B A22B Thinking
Intelligence Index28.5✓27.6
Coding Index31.9✓20.9
GPQA Diamond75%77%✓
Design Arena Elo1238✓—
Speed (tokens/sec)2645✓
Latency933ms✓1.0s
Input price /M$0.270$0.260✓
Output price /M$0.950✓$2.60
Context window164K✓131K
CapabilitiesReasoningToolsJSONReasoningToolsJSONVision
DeepSeek: DeepSeek V3.1 Terminus details →Qwen: Qwen3 VL 235B A22B Thinking details →