modelgrep

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

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

DeepSeek: DeepSeek V3.1 wins on more metrics (8 of 9), but the right pick depends on what you optimize for — see the breakdown below.

MetricDeepSeek: DeepSeek V3.1Qwen: Qwen3 VL 235B A22B Thinking
Intelligence Index28.1✓27.6
Coding Index28.4✓20.9
GPQA Diamond74%77%✓
Design Arena Elo1167✓—
Speed (tokens/sec)96✓45
Latency399ms✓1.0s
Input price /M$0.210✓$0.260
Output price /M$0.790✓$2.60
Context window164K✓131K
CapabilitiesReasoningToolsJSONReasoningToolsJSONVision
DeepSeek: DeepSeek V3.1 details →Qwen: Qwen3 VL 235B A22B Thinking details →