DeepSeek V4 Pro 0423 vs Kimi K2 Thinking
DeepSeek V4 Pro 0423 edges ahead on overall intelligence (RunFree 72.4 vs 45.2). Here's how they stack up on benchmarks, price and specs.
DeepSeek V4 Pro 0423
DeepSeek
Kimi K2 Thinking
MoonshotAI
Shared benchmarks (6)
Loading chart…
| DeepSeek V4 Pro 0423 | Kimi K2 Thinking | |
|---|---|---|
| RunFree Score | 72.4 | 45.2 |
| Blended price / 1M | $0.43 | $1.07 |
| Input / 1M | $0.35 | $0.60 |
| Output / 1M | $0.70 | $2.50 |
| Context window | 1.0M | 262K |
| Max output | 384K | 98K |
| Reasoning model | Yes | Yes |
| GPQA Diamond | 90.1% | 84.5% |
| Humanity's Last Exam | 37.7% | 23.9% |
| LiveCodeBench | 93.5% | 83.1% |
| MMLU-Pro | 87.5% | 84.6% |
| SWE-Bench Verified | 80.6% | 71.3% |
| Terminal-Bench | 67.9% | 47.1% |
| Wins | 12 | 0 |
A blank means the metric is tied or one model has no verified data. Prices are blended 3:1 input:output.
Compare more: DeepSeek V4 Pro 0423 card · Kimi K2 Thinking card · full leaderboard