modelgrep

DeepSeek: R1 0528 vs Qwen: Qwen3 VL 235B A22B Thinking

DDeepSeek: R1 0528deepseek/deepseek-r1-0528VSQQwen: Qwen3 VL 235B A22B Thinkingqwen/qwen3-vl-235b-a22b-thinking

DeepSeek: R1 0528 wins on more metrics (6 of 9), but the right pick depends on what you optimize for — see the breakdown below.

MetricDeepSeek: R1 0528Qwen: Qwen3 VL 235B A22B Thinking
Intelligence Index27.127.6✓
Coding Index24.0✓20.9
GPQA Diamond81%✓77%
Design Arena Elo1222✓—
Speed (tokens/sec)2645✓
Latency648ms✓1.0s
Input price /M$0.500$0.260✓
Output price /M$2.15✓$2.60
Context window164K✓131K
CapabilitiesReasoningToolsJSONReasoningToolsJSONVision
DeepSeek: R1 0528 details →Qwen: Qwen3 VL 235B A22B Thinking details →