FreeLLMAll models › Compare
DeepSeek-V4-Pro vs Kimi K2.5 (Thinking)
DeepSeek-V4-Pro and Kimi K2.5 (Thinking) are about even on Arena (1451 vs 1445): in a blind vote each wins about half the time.
DeepSeek-V4-Pro1451Arena chat · #49 of 417
Kimi K2.5 (Thinking)1445Arena chat · #61 of 417
Arena scores by category
| Category | DeepSeek-V4-Pro | Kimi K2.5 (Thinking) | Ahead |
|---|---|---|---|
| Chat | 1451 | 1445 | about even |
| Coding | 1582 | 1437 | DeepSeek-V4-Pro |
Scores from blind human votes on Arena. A gap under 10 points is too small to call.
Size and where to use them free
| DeepSeek-V4-Pro | Kimi K2.5 (Thinking) | |
|---|---|---|
| Made by | deepseek | moonshot |
| Size | 1598.8 B (75.6 B active) | 1026.9 B (33.3 B active) |
| 4-bit download | 970 GB | 624 GB |
| Use it free | download and run at home · MIT license | download and run at home · Modified MIT license |
Speed at home
| Hardware | DeepSeek-V4-Pro | Kimi K2.5 (Thinking) |
|---|---|---|
| RTX 3090 | does not fit | does not fit |
| RTX 4090 | does not fit | does not fit |
| RTX 5090 | does not fit | does not fit |
| RX 7900 XTX | does not fit | does not fit |
| Mac M4 Max 128 GB | does not fit | does not fit |
| Ryzen AI Max+ 395 128 GB | does not fit | does not fit |
Estimates at 4-bit with an 8K context, 64 GB of system RAM for offload, calibrated on 44 real llama.cpp tests.
Try them yourself
Ask both the same question side by side in Try & compare, or vote blind between free models in Free Arena.
More comparisons
- DeepSeek-V4-Pro vs MiMo-V2.6-Pro
- DeepSeek-V4-Pro vs Kimi K3 (Max)
- DeepSeek-V4-Pro vs GLM-5.3 (Max)
- DeepSeek-V4-Pro vs GLM-5.3-Flash
- DeepSeek-V4-Pro vs GLM-5.2 (Max)
- DeepSeek-V4-Pro vs MiMo-V2.5-Pro
- Kimi K2.5 (Thinking) vs MiMo-V2.6-Pro
- Kimi K2.5 (Thinking) vs Kimi K3 (Max)
- Kimi K2.5 (Thinking) vs GLM-5.3 (Max)
- Kimi K2.5 (Thinking) vs GLM-5.3-Flash
Share this page:XLinkedInRedditWhatsApp
Spotted a wrong number or a broken link? Report a mistake