Novita: Models Intelligence, Performance & Price
Analysis of Novita'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 Novita for your use-case. For more details including relating to our methodology, see our FAQs. Models analyzed: Llama 3.3 70B, Llama 4 Maverick (FP8), Llama 4 Scout, DeepSeek R1 Turbo, DeepSeek R1, DeepSeek R1 Distill Llama 70B, DeepSeek V3 (Mar' 25), DeepSeek R1 Distill Llama 8B, DeepSeek R1 Distill Qwen 32B, DeepSeek R1 Distill Qwen 14B, Qwen3 235B A22B (Reasoning) (FP8), Qwen3 30B A3B (Reasoning) (FP8), Qwen3 8B (Reasoning) (FP8), Qwen3 1.7B (Reasoning) (FP8), Qwen3 4B (Reasoning) (FP8), Qwen3 14B (Reasoning) (FP8), Qwen3 32B (Reasoning) (FP8), Llama 3.1 70B, Llama 3.1 8B, Llama 3.2 3B, Llama 3 70B, Llama 3 8B, Mistral 7B, DeepSeek V3 (Dec '24) Turbo, and DeepSeek V3 (Dec '24).
Link:
Novita Model Comparison Summary
Intelligence:
Qwen3 235B A22B (Reasoning) (FP8) and 
DeepSeek R1 Turbo are the highest quality models offered by Novita, followed by DeepSeek R1, Qwen3 32B (Reasoning) (FP8) & Qwen3 14B (Reasoning) (FP8).Output Speed (tokens/s): Qwen3 1.7B (Reasoning) (FP8) (168 t/s) and Qwen3 4B (Reasoning) (FP8) (128 t/s) are the fastest models offered by Novita, followed by 
Mistral 7B, 
Llama 3.2 3B & Llama 4 Scout.Latency (seconds): Llama 4 Maverick (FP8) (0.57s) and Llama 3.2 3B (0.58s) are the lowest latency models offered by Novita, followed by Qwen3 1.7B (Reasoning) (FP8), Llama 3.3 70B & Llama 4 Scout.Blended Price ($/M tokens): Qwen3 1.7B (Reasoning) (FP8) ($0.00) and Qwen3 4B (Reasoning) (FP8) ($0.00) are the cheapest models offered by Novita, followed by Llama 3.1 8B, Llama 3.2 3B & Mistral 7B.Context Window Size: Llama 4 Maverick (FP8) (1m) and Llama 4 Scout (131k) are the largest context window models offered by Novita, followed by Llama 3.3 70B, DeepSeek V3 (Mar' 25) & Qwen3 235B A22B (Reasoning) (FP8).
Qwen3 235B A22B (Reasoning) (FP8) and DeepSeek R1 Turbo are the highest quality models offered by Novita, followed by DeepSeek R1, Qwen3 32B (Reasoning) (FP8) & Qwen3 14B (Reasoning) (FP8).Output Speed (tokens/s): Qwen3 1.7B (Reasoning) (FP8) (168 t/s) and Qwen3 4B (Reasoning) (FP8) (128 t/s) are the fastest models offered by Novita, followed by Mistral 7B, Llama 3.2 3B & Llama 4 Scout.Latency (seconds): Llama 4 Maverick (FP8) (0.57s) and Llama 3.2 3B (0.58s) are the lowest latency models offered by Novita, followed by Qwen3 1.7B (Reasoning) (FP8), Llama 3.3 70B & Llama 4 Scout.Blended Price ($/M tokens): Qwen3 1.7B (Reasoning) (FP8) ($0.00) and Qwen3 4B (Reasoning) (FP8) ($0.00) are the cheapest models offered by Novita, followed by Llama 3.1 8B, Llama 3.2 3B & Mistral 7B.Context Window Size: Llama 4 Maverick (FP8) (1m) and Llama 4 Scout (131k) are the largest context window models offered by Novita, followed by Llama 3.3 70B, DeepSeek V3 (Mar' 25) & Qwen3 235B A22B (Reasoning) (FP8).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 | End-to-End Response Time | ||||
|---|---|---|---|---|---|---|---|---|---|
Further Analysis | |||||||||
 | Qwen3 235B A22B (Reasoning) (FP8) | 128k | 62 | $0.35 | 33.2 | 0.71 | 75.94 | 60.18 | |
| DeepSeek R1 Turbo | 64k | 60 | $1.15 | 33.1 | 0.80 | 86.73 | 70.83 | |
| DeepSeek R1 | 64k | 60 | $4.00 | 34.1 | 1.06 | 84.46 | 68.76 | |
| Qwen3 32B (Reasoning) (FP8) | 128k | 59 | $0.19 | 27.7 | 0.89 | 91.25 | 72.29 | |
| Qwen3 14B (Reasoning) (FP8) | 128k | 56 | $0.12 | 45.2 | 0.79 | 56.03 | 44.20 | |
| Qwen3 30B A3B (Reasoning) (FP8) | 128k | 56 | $0.19 | 67.8 | 0.67 | 37.55 | 29.50 | |
| DeepSeek V3 (Mar' 25) | 128k | 53 | $0.57 | 27.7 | 0.92 | 18.98 | N/A | |
| DeepSeek R1 Distill Qwen 32B | 64k | 52 | $0.30 | 20.7 | 1.14 | 121.63 | 96.40 | |
| Qwen3 8B (Reasoning) (FP8) | 128k | 51 | $0.06 | 68.9 | 0.72 | 37.02 | 29.04 | |
| Llama 4 Maverick (FP8) | 1m | 51 | $0.34 | 73.7 | 0.57 | 7.35 | N/A | |
| DeepSeek R1 Distill Qwen 14B | 64k | 49 | $0.15 | 24.2 | 0.89 | 104.12 | 82.58 | |
| DeepSeek R1 Distill Llama 70B | 32k | 48 | $0.39 | 72.9 | 0.64 | 34.93 | 27.43 | |
| Qwen3 4B (Reasoning) (FP8) | 128k | 47 | $0.00 | 127.8 | 0.64 | 20.21 | 15.65 | |
| DeepSeek V3 (Dec '24) Turbo | 64k | 46 | $0.63 | 30.5 | 0.83 | 17.24 | N/A | |
| DeepSeek V3 (Dec '24) | 64k | 46 | $0.89 | 30.6 | 0.84 | 17.17 | N/A | |
| Llama 4 Scout | 131k | 43 | $0.20 | 76.4 | 0.62 | 7.16 | N/A | |
| Llama 3.3 70B | 128k | 41 | $0.20 | 74.9 | 0.62 | 7.29 | N/A | |
| Qwen3 1.7B (Reasoning) (FP8) | 32k | 38 | $0.00 | 167.7 | 0.60 | 15.51 | 11.93 | |
| Llama 3.1 70B | 32k | 35 | $0.19 | 54.9 | 1.06 | 10.16 | N/A | |
| DeepSeek R1 Distill Llama 8B | 32k | 34 | $0.04 | 51.7 | 0.72 | 49.12 | 38.72 | |
| Llama 3 70B | 8k | 27 | $0.57 | 20.0 | 1.00 | 26.00 | N/A | |
| Llama 3.1 8B | 16k | 24 | $0.03 | 74.8 | 0.67 | 7.35 | N/A | |
| Llama 3 8B | 8k | 21 | $0.04 | 58.6 | 0.85 | 9.38 | N/A | |
| Llama 3.2 3B | 32k | 20 | $0.04 | 109.9 | 0.58 | 5.13 | N/A | |
| Mistral 7B | 32k | 10 | $0.04 | 122.4 | 0.80 | 4.89 | N/A | |
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): Time to first token received, in seconds, after API request sent. For reasoning models which share reasoning tokens, this will be the first reasoning token. 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.