Inference provider
Nebius
10 models served. Speed, latency and price below are medians across this provider's catalogue, not per-endpoint measurements.
Models served
10
Median output speed
88 t/s
Median latency
378ms
Median price / 1M
$0.234
Catalogue
Strongest and fastest on this provider
What the catalogue looks like at the top end, on the two axes that usually decide the choice.
Highest Intelligence Index
Among models this provider serves
- GLM 5.2Z.ai33.7
- GLM 5.1Z.ai26.1
- Kimi K2.7 CodeMoonshot AI25.8
- gpt-oss-120bOpenAI11.6
- Nemotron 3 Nano 30B A3BNVIDIA6.8
View as table
| Model | Intelligence |
|---|---|
| GLM 5.2 (Z.ai) | 33.7 |
| GLM 5.1 (Z.ai) | 26.1 |
| Kimi K2.7 Code (Moonshot AI) | 25.8 |
| gpt-oss-120b (OpenAI) | 11.6 |
| Nemotron 3 Nano 30B A3B (NVIDIA) | 6.8 |
Fastest output
Median tokens per second
- gpt-oss-120bOpenAI641 t/s
- GLM 5.2Z.ai368 t/s
- Nemotron 3 Nano 30B A3BNVIDIA201 t/s
- Kimi K2.7 CodeMoonshot AI127 t/s
- GLM 5.1Z.ai100 t/s
- Hermes 4 70BNous Research76 t/s
- 74 t/s
- 49 t/s
- Hermes 4 405BNous Research37 t/s
- Gemma 3 27BGoogle28 t/s
View as table
| Model | Tokens/s |
|---|---|
| gpt-oss-120b (OpenAI) | 641 |
| GLM 5.2 (Z.ai) | 368 |
| Nemotron 3 Nano 30B A3B (NVIDIA) | 201 |
| Kimi K2.7 Code (Moonshot AI) | 127 |
| GLM 5.1 (Z.ai) | 100 |
| Hermes 4 70B (Nous Research) | 76 |
| Qwen3 30B A3B Instruct 2507 (Qwen) | 74 |
| Qwen3 235B A22B Instruct 2507 (Qwen) | 49 |
| Hermes 4 405B (Nous Research) | 37 |
| Gemma 3 27B (Google) | 28 |
Full catalogue
All 10 models on Nebius
Filter by creator
10 of 10 models| GLM 5.2 Z.aiopen | 33.7 | $0.56 |
| GLM 5.1 Z.aiopen | 26.1 | $0.29 |
| Kimi K2.7 Code Moonshot AIopen | 25.8 | $0.30 |
| gpt-oss-120b OpenAIopen | 11.6 | $0.06 |
| Nemotron 3 Nano 30B A3B NVIDIAopen | 6.8 | $0.02 |
| Hermes 4 70B Nous Researchopen | — | $0.04 |
| Hermes 4 405B Nous Researchopen | — | $0.29 |
| Qwen3 30B A3B Instruct 2507 Qwenopen | — | $0.01 |
| Qwen3 235B A22B Instruct 2507 Qwenopen | — | $0.02 |
| Gemma 3 27B Googleopen | — | $0.02 |