Nebius: Models Intelligence, Performance & Price
Analysis of Nebius's models across key metrics including quality, price, output speed, latency, context window & more. This analysis is intended to support you in choosing the best model provided by Nebius for your use-case. For more details including relating to our methodology, see our FAQs. Models analyzed: Llama 3.3 70B Fast, Llama 3.3 70B Base, Llama 3.1 405B Base, Llama 3.1 70B Base, Llama 3.1 70B Fast, Llama 3.1 8B Fast, Llama 3.1 8B Base, Llama 3.2 3B Base, Llama 3.2 1B Base, Gemma 2 27B Fast, Gemma 2 27B Base, Gemma 2 9B Fast, Gemma 2 9B Base, Mixtral 8x22B Base, Mixtral 8x22B Fast, Mistral NeMo Fast, Mistral NeMo Base, Mixtral 8x7B Fast, Mixtral 8x7B Base, DeepSeek R1 Base, DeepSeek R1 Fast, DeepSeek R1 Distill Llama 70B Base, DeepSeek V3, DeepSeek Coder V2 Lite Fast, FP8, DeepSeek Coder V2 Lite Base, FP8, Phi-4, Llama 3.1 Nemotron 70B Base, Llama 3.1 Nemotron 70B Fast, Qwen2.5 72B, Qwen2.5 72B Fast, QwQ-32B Fast, QwQ-32B Base, Qwen2.5 Instruct 32B Fast, Qwen2.5 Instruct 32B Base, Qwen2.5 Coder 7B Fast, Qwen2.5 Coder 7B Base, and QwQ 32B-Preview.
Link:
Nebius Model Comparison Summary
Intelligence:
DeepSeek R1 Base and DeepSeek R1 Fast are the highest quality models offered by Nebius, followed by 
QwQ-32B Fast, QwQ-32B Base & DeepSeek R1 Distill Llama 70B Base.Output Speed (tokens/s):
Llama 3.2 1B Base (267 t/s) and Qwen2.5 Coder 7B Fast (220 t/s) are the fastest models offered by Nebius, followed by Qwen2.5 Coder 7B Base, Llama 3.1 8B Fast & 
Gemma 2 9B Base.Latency (seconds): Llama 3.2 1B Base (0.46s) and Llama 3.1 8B Fast (0.47s) are the lowest latency models offered by Nebius, followed by Qwen2.5 Coder 7B Base, 
Mistral NeMo Fast & Gemma 2 9B Base.Blended Price ($/M tokens): Llama 3.2 1B Base ($0.01) and Llama 3.2 3B Base ($0.01) are the cheapest models offered by Nebius, followed by Qwen2.5 Coder 7B Base, Llama 3.1 8B Base & Gemma 2 9B Base.Context Window Size: Qwen2.5 72B (131k) and Qwen2.5 72B Fast (131k) are the largest context window models offered by Nebius, followed by QwQ-32B Fast, QwQ-32B Base & Qwen2.5 Coder 7B Fast.
DeepSeek R1 Base and DeepSeek R1 Fast are the highest quality models offered by Nebius, followed by QwQ-32B Fast, QwQ-32B Base & DeepSeek R1 Distill Llama 70B Base.Output Speed (tokens/s): Llama 3.2 1B Base (267 t/s) and Qwen2.5 Coder 7B Fast (220 t/s) are the fastest models offered by Nebius, followed by Qwen2.5 Coder 7B Base, Llama 3.1 8B Fast & Gemma 2 9B Base.Latency (seconds): Llama 3.2 1B Base (0.46s) and Llama 3.1 8B Fast (0.47s) are the lowest latency models offered by Nebius, followed by Qwen2.5 Coder 7B Base, Mistral NeMo Fast & Gemma 2 9B Base.Blended Price ($/M tokens): Llama 3.2 1B Base ($0.01) and Llama 3.2 3B Base ($0.01) are the cheapest models offered by Nebius, followed by Qwen2.5 Coder 7B Base, Llama 3.1 8B Base & Gemma 2 9B Base.Context Window Size: Qwen2.5 72B (131k) and Qwen2.5 72B Fast (131k) are the largest context window models offered by Nebius, followed by QwQ-32B Fast, QwQ-32B Base & Qwen2.5 Coder 7B Fast.Highlights
Intelligence
Artificial Analysis Intelligence Index; Higher is better
Speed
Output Tokens per Second; Higher is better
Price
USD per 1M Tokens; Lower is better
Parallel Queries:
Prompt Length:
Features | Model Intelligence | Price | Output tokens/s | Latency | |||
|---|---|---|---|---|---|---|---|
Further Analysis | |||||||
 | DeepSeek R1 Base | 128k | 60 | $1.20 | 9.3 | 0.93 | |
| DeepSeek R1 Fast | 128k | 60 | $3.00 | 66.3 | 0.66 | |
| QwQ-32B Fast | 131k | 58 | $0.75 | 72.1 | 0.51 | |
| QwQ-32B Base | 131k | 58 | $0.23 | 19.9 | 0.63 | |
| DeepSeek R1 Distill Llama 70B Base | 128k | 48 | $0.38 | 32.5 | 0.59 | |
| QwQ 32B-Preview | 33k | 43 | $0.14 | 59.7 | 0.64 | |
| Llama 3.3 70B Fast | 128k | 41 | $0.38 | 133.9 | 0.54 | |
| Llama 3.3 70B Base | 128k | 41 | $0.20 | 19.1 | 0.68 | |
| Qwen2.5 72B | 131k | 41 | $0.20 | 23.0 | 0.70 | |
| Qwen2.5 72B Fast | 131k | 41 | $0.38 | 70.4 | 0.56 | |
| Llama 3.1 405B Base | 128k | 40 | $1.50 | 34.7 | 0.68 | |
|  Phi-4 | 16k | 40 | $0.15 | 117.5 | 0.52 | |
| Qwen2.5 Instruct 32B Fast | 128k | 37 | $0.20 | 85.7 | 0.55 | |
| Qwen2.5 Instruct 32B Base | 128k | 37 | $0.10 | 60.7 | 0.56 | |
|  Llama 3.1 Nemotron 70B Base | 128k | 37 | $0.20 | 38.3 | 0.65 | |
| Llama 3.1 Nemotron 70B Fast | 128k | 37 | $0.38 | 73.3 | 0.53 | |
| Llama 3.1 70B Base | 128k | 35 | $0.20 | 28.5 | 0.67 | |
| Llama 3.1 70B Fast | 128k | 35 | $0.38 | 147.3 | 0.56 | |
| Mixtral 8x22B Base | 65k | 26 | $0.60 | 91.0 | 0.50 | |
| Mixtral 8x22B Fast | 65k | 26 | $1.05 | 108.1 | 0.53 | |
| Llama 3.1 8B Fast | 128k | 24 | $0.04 | 183.7 | 0.47 | |
| Llama 3.1 8B Base | 128k | 24 | $0.03 | 66.4 | 0.53 | |
| Mistral NeMo Fast | 128k | 20 | $0.12 | 159.6 | 0.49 | |
| Mistral NeMo Base | 128k | 20 | $0.06 | 22.0 | 0.71 | |
| Llama 3.2 3B Base | 128k | 20 | $0.01 | 122.5 | 0.51 | |
| Mixtral 8x7B Fast | 33k | 17 | $0.23 | 160.7 | 0.50 | |
| Mixtral 8x7B Base | 33k | 17 | $0.12 | 54.1 | 0.60 | |
| Llama 3.2 1B Base | 128k | 10 | $0.01 | 267.0 | 0.46 | |
| Gemma 2 27B Fast | 8k | $0.26 | 86.0 | 0.52 | ||
| Gemma 2 27B Base | 8k | $0.15 | 51.7 | 0.56 | ||
| Gemma 2 9B Fast | 8k | $0.04 | 162.6 | 0.50 | ||
| Gemma 2 9B Base | 8k | $0.03 | 166.4 | 0.49 | ||
| DeepSeek V3 | 128k | $0.75 | 23.9 | 0.93 | ||
| DeepSeek Coder V2 Lite Fast, FP8 | 128k | $0.12 | 115.7 | 0.58 | ||
| DeepSeek Coder V2 Lite Base, FP8 | 128k | $0.06 | 110.2 | 0.57 | ||
| Qwen2.5 Coder 7B Fast | 131k | $0.04 | 219.8 | 0.50 | ||
| Qwen2.5 Coder 7B Base | 131k | $0.01 | 193.1 | 0.48 |
Key definitions
Context window: Maximum number of combined input & output tokens. Output tokens commonly have a significantly lower limit (varied by model).
Output Speed: Tokens per second received while the model is generating tokens (ie. after first chunk has been received from the API for models which support streaming).
Latency: Time to first token of tokens received, in seconds, after API request sent. For models which do not support streaming, this represents time to receive the completion.
Price: Price per token, represented as USD per million Tokens. Price is a blend of Input & Output token prices (3:1 ratio).
Output Price: Price per token generated by the model (received from the API), represented as USD per million Tokens.
Input Price: Price per token included in the request/message sent to the API, represented as USD per million Tokens.
Time period: Metrics are 'live' and are based on the past 72 hours of measurements, measurements are taken 8 times a day for single requests and 2 times per day for parallel requests.