Compare / DeepSeek V3.2 vs Kimi K2 Thinking
DeepSeek V3.2 vs Kimi K2 Thinking
16 results from 3 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | DeepSeek V3.2 | Kimi K2 Thinking |
|---|---|---|---|
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 21.5 | 22.0 |
| Artificial Analysisreported | GPQA Diamond% of questions | 84.0% | 83.8% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 24.6% | 23.8% |
| Artificial Analysisreported | IFBench% of instructions | 60.7% | 68.1% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 73.3% | 72.0% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 35.6% | 31.1% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 90.6% | 93.0% |
| OpenHands Indexreported | Average% resolved | 45.7% | 41.0% |
| OpenHands Indexreported | Front end (SWE-Bench Multimodal)% resolved | 27.9% | 32.4% |
| OpenHands Indexreported | Greenfield (Commit0)% resolved | 25.0% | 12.5% |
| OpenHands Indexreported | Information gathering (GAIA)% resolved | 50.3% | 43.6% |
| OpenHands Indexreported | Issue resolution (SWE-Bench)% resolved | 71.6% | 69.2% |
| OpenHands Indexreported | Testing (SWT-Bench)% resolved | 53.6% | 47.3% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 26.0% | 20.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.36 | 0.38 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 57.3 | 59.3 |
Bold green marks the better value on that metric. Where a source reports ranges that overlap, the difference may not be meaningful; see the benchmark page for ranges.