DeepSeek V3.2 vs Kimi K2.5
Head to head on 25 shared benchmarks, with price, context window, and release dates.
Benchmark scores
45evals · sort by any column, ⤢ to expand fullscreen
| Model | ||||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Kimi K2.5 Moonshot AI | 68.3% | 39.0% | 1576elo | 74.9 | 27.9% | 78.9% | 13.2% | 43.7% | 78.5% | 72.5% | 77.9% | 61.4% | 57.4% | 77.7% | 84.9% | 76.0% | 62.0% | 39.6% | 67.3% | 70.8% | 74.2% | 18.9% | 69.4 | 17.5% | 81.3% | - | - | - | 96.1 | - | 35.8% | 95.4 | - | 10421 | - | 39.3% | 76.4% | 87.1 | - | 66.5% | 63.3% | - | 49.9% | 73 | 26.3% | 56.8% |
| DeepSeek V3.2 DeepSeek | 47.9% | 16.0% | 1511elo | 66.8 | 22.1% | 75.1% | 11.2% | 49.0% | 68.0% | 65.9% | 64.6% | 50.0% | 48.2% | 70.4% | 85.0% | 77.2% | 57.5% | 38.7% | 59.0% | 70.0% | 68.2% | 32.6% | 65 | 5.1% | 78.9% | 74.7% | 74.2% | 59.0% | - | 0.0% | - | - | 31.3% | - | 59.3% | - | - | - | 83.7% | - | - | 8.0% | - | - | - | 51.7% |
2 / 2 models
| Attribute | DeepSeek V3.2 DeepSeek | Kimi K2.5 Moonshot AI |
|---|---|---|
| Description | DeepSeek V3.2 is an AI model from DeepSeek, released with open weights. | Kimi K2.5 is an AI model from Kimi. |
| Family | Deepseek | Kimi |
| Params | - | - |
| Open weights | open | closed |
| License | mit | Modified MIT |
| Released | 1 Dec 2025 | 27 Jan 2026 |
| Context | 163,840 | 262,144 |
| Input $/1M | $0.28 | $0.60 |
| Output $/1M | $0.42 | $3.00 |
| Cache read $/1M | $0.130 | $0.100 |
| Cache write $/1M | - | - |
| Speed | 0 tok/s | 0 tok/s |
| Latency | 0.00 s | 0.00 s |
| Modalities | text | text, image |
| Providers | - | - |
| Publisher | DeepSeek | Moonshot AI |
| Reported scores | Aider Polyglot Benchmark74.2% (Aider) MathArena57.47% GPQA Diamond75.1% (AA) Humanity's Last Exam (HLE)11.2% (AA) | GPQA Diamond78.9% (AA) Humanity's Last Exam (HLE)13.2% (AA) IFBench43.7% (AA) SciCode39.6% (AA) τ²-bench (Tau²-bench)81.3% (AA) |
FAQ
- Is DeepSeek V3.2 or Kimi K2.5 better?
- Across the 25 benchmarks both models report on Sophon, Kimi K2.5 leads on 21 of them. Which one is "better" depends on the benchmark - the table above shows every shared score.
- How do DeepSeek V3.2 and Kimi K2.5 compare on GPQA Diamond?
- DeepSeek V3.2 scores 75.1% and Kimi K2.5 scores 78.9% on GPQA Diamond.
- Which is cheaper, DeepSeek V3.2 or Kimi K2.5?
- DeepSeek V3.2 is cheaper at $0.42 per million output tokens against $3.00 for Kimi K2.5 - about 7.1x.
- When were DeepSeek V3.2 and Kimi K2.5 released?
- DeepSeek V3.2 was released 1 Dec 2025 by DeepSeek. Kimi K2.5 was released 27 Jan 2026 by Moonshot AI.
Comparing something else? Build your own side-by-side.