FreeLLMAll models › Compare
Kimi K2.5 (Thinking) vs NVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4)
Kimi K2.5 (Thinking) is ahead on Arena (1450 vs 1427): in a blind vote people prefer it about 53% of the time.
Kimi K2.5 (Thinking)1450Arena chat · #80 of 419
NVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4)1427Arena chat · #120 of 419
Arena scores by category
| Category | Kimi K2.5 (Thinking) | NVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4) | Ahead |
|---|---|---|---|
| Chat | 1450 | 1427 | Kimi K2.5 (Thinking) |
Scores from blind human votes on Arena. A gap under 10 points is too small to call.
Size and where to use them free
| Kimi K2.5 (Thinking) | NVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4) | |
|---|---|---|
| Made by | moonshot | nvidia |
| Size | 1026.9 B (32.2 B active) | 560.5 B (55 B active) |
| 4-bit download | 624 GB | 341 GB |
| Use it free | download and run at home · Modified MIT license | free API on OpenRouter · download and run at home · OpenMDW-1.1 license |
Speed at home
| Hardware | Kimi K2.5 (Thinking) | NVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4) |
|---|---|---|
| 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.
Share this page:XLinkedInRedditWhatsApp
Spotted a wrong number or a broken link? Report a mistake