Novita: Models Intelligence, Performance & Price
This analysis is intended to support you in choosing the best model provided by Novita for your use-case.
Most Intelligent
DeepSeek V4.1 Flash 1017 (max)
GLM-5.3 (max)
MiniMax-M3.1-FlashGLM-5.3-FlashDeepSeek V4 Pro 0813 (max)Intelligence Index
Total 61 models
Lowest Time per Task
gpt-oss-120b (high)MiniMax-M3GLM-5.3 (max)MiniMax-M3.1-FlashDeepSeek V4.1 Flash 1017 (max)Time per Intelligence Index task (minutes)
Total 61 models
Lowest Price
Llama 3.1 8BGLM-4.7-FlashGLM-4.7-Flash (non-reasoning)
Nemotron 3 Nano (non-reasoning) (FP4)Nemotron 3 Nano (FP4)Blended price (per 1M tokens)
Total 61 models
Novita offers 61 models, each with different intelligence, performance, and pricing characteristics. Below is a comparison of the key metrics across models.
- For intelligence, the top models on Novita are DeepSeek V4.1 Flash 1017 (max) (33), GLM-5.3 (max) (32), and MiniMax-M3.1-Flash (29).
- For time per task, gpt-oss-120b (high) (8.0 min), MiniMax-M3 (14.8 min), and GLM-5.3 (max) (18.2 min) finish an Intelligence Index task fastest on Novita. Time per task varies significantly across models, with the slowest taking 2.4x longer than the quickest.
- For latency, Gemma 3 27B (0.86s), DeepSeek V3 (Dec) (1.50s), and Llama 4 Scout (1.61s) offer the lowest time to first answer token.
- For pricing, Llama 3.1 8B ($0.02), GLM-4.7-Flash ($0.06), and GLM-4.7-Flash (non-reasoning) ($0.06) offer the lowest blended prices per 1M tokens. Prices vary up to 2.8x across models.
- For context window size, DeepSeek V4.1 Flash 1017 (max) (1M), GLM-5.3-Flash (1M), and DeepSeek V4 Pro 0813 (max) (1M) support the largest context windows on Novita.
Intelligence
Artificial Analysis Intelligence Index
Evaluations
Agentic knowledge work, Elo
Knowledge
1 - hallucination rate
Agentic SaaS workflows
Agentic scientific research workflows in a terminal
Agentic software engineering in open-source repositories
Reasoning & knowledge
Professional document reasoning, All-pass
Agentic coding & terminal use
Physics reasoning
Quantitative analysis on spreadsheets & documents
Legal agentic work, Hallucination-Gated All-Pass Rate
Agentic business operations
Agentic real-world work tasks, Elo
Agentic interactive 3D experiences, Elo
Product and brand videos designed in code, Elo
Kubernetes incident root-cause analysis
Medical long context reasoning
Visual reasoning
Rich, varied, well-crafted language
Well-structured, scannable, answer-first
Truthful, acknowledges uncertainty, non-sycophantic
Use of filler and generic language
Intelligence Index vs. Price
Pricing
Cost per Intelligence Index Task
Intelligence Index vs. Price
Speed
Average of the single-request and multi-turn agentic workloads measured
Time per Intelligence Index Task
Quality
Endpoint Accuracy Index
Latency
Average of the single-request and multi-turn agentic workloads measured
Time to First Answer Token
Reliability
Share of requests the endpoint failed, across all the single-request and multi-turn agentic workloads measured
Endpoint Error Rate
Caching
Cache Hit Rate
Cost per Task vs. Cache Hit Rate
Further Analysis | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
DeepSeek V4.1 Flash 1017 (max) | 33 | $0.52 | 19.0 | 57.10 | 1.0% | 1M | Open | |||
GLM-5.3 (max) | 32 | $5.00 | 17.5 | 58.89 | 0.0% | 1M | Open | |||
MiniMax-M3.1-Flash | 29 | $0.43 | 17.0 | 56.86 | 0.0% | 1M | Open | |||
GLM-5.3-Flash | 28 | $0.80 | 29.2 | 63.08 | 0.0% | 1M | Open | |||
DeepSeek V4 Pro 0813 (max) | 23 | $1.86 | 25.2 | 79.39 | 0.0% | 1M | Open | |||
MiniMax-M3 | 19 | $1.16 | 13.3 | 67.89 | 0.0% | 1M | Open | |||
 Gemma 4 31B | 4 | -- | -- | 62.27 | 0.0% | 262k | Open | |||
gpt-oss-120b (high) | 4 | $0.12 | 7.4 | 13.77 | 2.5% | 131k | Open | |||
GLM-4.5V (non-reasoning) | -- | -- | -- | 4.87 | 0.0% | 66k | Open | |||
DeepSeek V3.1 Terminus (FP8) | -- | -- | -- | -- | -- | 131k | Open | |||
 Kimi K2.6 | -- | -- | -- | 66.81 | 0.0% | 262k | Open | |||
DeepSeek R1 (Jan) Turbo | -- | -- | -- | -- | -- | 64k | Open | |||
DeepSeek R1 (Jan) | -- | -- | -- | -- | -- | 64k | Open | |||
 Qwen3.5 397B A17B | -- | -- | -- | 40.01 | 1.0% | 262k | Open | |||
Kimi K2.6 (non-reasoning) | -- | -- | -- | 7.40 | 0.0% | 262k | Open | |||
MiniMax-M2.5 | -- | -- | -- | 19.64 | 0.0% | 205k | Open | |||
Qwen3.5 27B | -- | -- | -- | 33.99 | 0.0% | 262k | Open | |||
DeepSeek V3.1 | -- | -- | -- | -- | -- | 131k | Open | |||
Qwen3.5 122B A10B | -- | -- | -- | 28.96 | 0.0% | 262k | Open | |||
Nemotron 3 Nano (non-reasoning) (FP4) | -- | -- | -- | 1.85 | 1.0% | 262k | Open | |||
Gemma 3 27B | -- | -- | -- | 0.91 | 0.0% | 98k | Open | |||
GLM-4.7-Flash | -- | -- | -- | 62.14 | 0.0% | 200k | Open | |||
GLM-4.5V | -- | -- | -- | 23.17 | 0.0% | 66k | Open | |||
Qwen3.5 397B A17B (non-reasoning) | -- | -- | -- | 2.80 | 0.0% | 262k | Open | |||
Qwen3.6 27B (non-reasoning) | -- | -- | -- | 36.38 | 5.3% | 262k | Open | |||
GLM-5 FP8 | -- | -- | -- | 20.41 | 0.0% | 203k | Open | |||
GLM-4.6V (non-reasoning) | -- | -- | -- | 8.74 | 0.0% | 131k | Open | |||
DeepSeek V3 (Dec) | -- | -- | -- | 1.25 | 0.0% | 64k | Open | |||
Kimi K2 | -- | -- | -- | 3.62 | 2.2% | 131k | Open | |||
Qwen3.5 35B A3B | -- | -- | -- | 36.31 | 0.0% | 262k | Open | |||
MiniMax-M2.7 (FP8) | -- | -- | -- | 25.72 | 0.0% | 205k | Open | |||
Qwen3 VL 30B A3B | -- | -- | -- | 2.37 | 0.0% | 131k | Open | |||
GLM-5.1 (FP8) | -- | -- | -- | 25.91 | 0.0% | 205k | Open | |||
DeepSeek V4 Pro (high) | -- | -- | -- | -- | -- | 1M | Open | |||
Kimi K2 0905 | -- | -- | -- | 2.83 | 0.0% | 262k | Open | |||
Qwen3.6 35B A3B (non-reasoning) | -- | -- | -- | 3.26 | 1.7% | 205k | Open | |||
GLM-4.6 | -- | -- | -- | 54.45 | 0.0% | 205k | Open | |||
Gemma 4 26B A4B | -- | -- | -- | 30.28 | 0.0% | 262k | Open | |||
GLM-4.7 | -- | -- | -- | 45.81 | 0.0% | 205k | Open | |||
Gemma 4 26B A4B (non-reasoning) | -- | -- | -- | 1.83 | 0.9% | 262k | Open | |||
Llama 3.1 8B | -- | -- | -- | 1.97 | 0.0% | 16k | Open | |||
Nemotron 3 Nano (FP4) | -- | -- | -- | 5.92 | 0.0% | 262k | Open | |||
GLM-4.7 (non-reasoning) | -- | -- | -- | 13.06 | 0.0% | 205k | Open | |||
Llama 3 8B | -- | -- | -- | -- | -- | 8k | Open | |||
Qwen3.6 35B A3B | -- | -- | -- | 18.42 | 1.1% | 205k | Open | |||
Qwen3.6 27B | -- | -- | -- | 46.63 | 3.3% | 262k | Open | |||
Kimi K2.7 Code | -- | -- | -- | 31.63 | 0.0% | 262k | Open | |||
Llama 4 Maverick (FP8) | -- | -- | -- | 2.78 | 0.0% | 1M | Open | |||
GLM-5.1 (non-reasoning) (FP8) | -- | -- | -- | 5.43 | 0.0% | 205k | Open | |||
GLM-4.6 (non-reasoning) | -- | -- | -- | 17.07 | 0.0% | 205k | Open | |||
Llama 3.3 70B | -- | -- | -- | 4.98 | 4.1% | 131k | Open | |||
MiniMax M1 80k | -- | -- | -- | 4.22 | 0.0% | 1M | Open | |||
Kimi K2.5 | -- | -- | -- | 31.90 | 0.0% | 262k | Open | |||
DeepSeek R1 0528 | -- | -- | -- | 80.00 | 2.5% | 164k | Open | |||
Llama 3 70B | -- | -- | -- | -- | -- | 8k | Open | |||
Llama 4 Scout | -- | -- | -- | 1.38 | 8.9% | 131k | Open | |||
 Hy3 | -- | -- | -- | 42.72 | 0.0% | 262k | Open | |||
Kimi K2.5 (non-reasoning) | -- | -- | -- | 5.86 | 0.0% | 262k | Open | |||
Kimi K2 Thinking | -- | -- | -- | 36.17 | 0.0% | 262k | Open | |||
DeepSeek V4 Pro (max) | -- | -- | -- | 94.29 | 0.0% | 1M | Open | |||
GLM-4.7-Flash (non-reasoning) | -- | -- | -- | 19.85 | 3.1% | 200k | Open | |||
gpt-oss-120b (low) | -- | -- | -- | 2.16 | 2.0% | 131k | Open | |||
Gemma 4 31B (non-reasoning) | -- | -- | -- | 7.66 | 0.8% | 262k | Open | |||
GLM-4.6V | -- | -- | -- | 19.94 | 0.0% | 131k | Open | |||
DeepSeek V4.1 Flash (max) | -- | -- | -- | 55.50 | 0.0% | 524k | Open | |||
DeepSeek V3.1 Terminus (non-reasoning) (FP8) | -- | -- | -- | -- | -- | 131k | Open | |||
GLM-5 (non-reasoning) (FP8) | -- | -- | -- | 4.74 | 0.0% | 203k | Open | |||
GLM-5.2 (max) (FP8) | -- | -- | -- | 28.45 | 0.0% | 1M | Open | |||
DeepSeek V3.1 (non-reasoning) | -- | -- | -- | -- | -- | 164k | Open | |||
DeepSeek V4 Flash 0731 (max) | -- | -- | -- | 48.45 | 0.0% | 1M | Open | |||
Key definitions
Frequently Asked Questions
Common questions about Novita
Novita offers 61 models that we track: DeepSeek V4.1 Flash 1017 (max), GLM-5.3 (max), MiniMax-M3.1-Flash, GLM-5.3-Flash, DeepSeek V4 Pro 0813 (max), MiniMax-M3, Gemma 4 31B, gpt-oss-120b (high), GLM-4.5V (non-reasoning), Kimi K2.6, Qwen3.5 397B A17B, Kimi K2.6 (non-reasoning), MiniMax-M2.5, Qwen3.5 27B, Qwen3.5 122B A10B, Nemotron 3 Nano (non-reasoning) (FP4), Gemma 3 27B, GLM-4.7-Flash, GLM-4.5V, Qwen3.5 397B A17B (non-reasoning), Qwen3.6 27B (non-reasoning), GLM-5 FP8, GLM-4.6V (non-reasoning), DeepSeek V3 (Dec), Kimi K2, Qwen3.5 35B A3B, MiniMax-M2.7 (FP8), Qwen3 VL 30B A3B, GLM-5.1 (FP8), Kimi K2 0905, Qwen3.6 35B A3B (non-reasoning), GLM-4.6, Gemma 4 26B A4B, GLM-4.7, Gemma 4 26B A4B (non-reasoning), Llama 3.1 8B, Nemotron 3 Nano (FP4), GLM-4.7 (non-reasoning), Qwen3.6 35B A3B, Qwen3.6 27B, Kimi K2.7 Code, Llama 4 Maverick (FP8), GLM-5.1 (non-reasoning) (FP8), GLM-4.6 (non-reasoning), Llama 3.3 70B, MiniMax M1 80k, Kimi K2.5, DeepSeek R1 0528, Llama 4 Scout, Hy3, Kimi K2.5 (non-reasoning), Kimi K2 Thinking, DeepSeek V4 Pro (max), GLM-4.7-Flash (non-reasoning), gpt-oss-120b (low), Gemma 4 31B (non-reasoning), GLM-4.6V, DeepSeek V4.1 Flash (max), GLM-5 (non-reasoning) (FP8), GLM-5.2 (max) (FP8), and DeepSeek V4 Flash 0731 (max).
The most intelligent model available on Novita is DeepSeek V4.1 Flash 1017 (max) with an Intelligence Index score of 33.
The model with the lowest time per task on Novita is gpt-oss-120b (high) at 8.0 minutes per Intelligence Index task. Time per task is the output tokens per task divided by end-to-end output speed, so lower means a task finishes sooner.
The model with the lowest time to first answer token on Novita is Gemma 3 27B at 0.86s. Lower latency means faster initial response time.
The most affordable model on Novita by blended price is Llama 3.1 8B at $0.023 per 1M tokens (7:2:1 cache hit/input/output ratio).
Prices on Novita vary up to 39.1x across models, from $0.023 per 1M tokens for Llama 3.1 8B to $0.90 per 1M tokens for GLM-5.2 (max) (FP8).
Yes, Novita offers an OpenAI-compatible API, making it easy to switch from OpenAI or use existing OpenAI SDK integrations.
60 of 61 models on Novita support JSON mode for structured output.
57 of 61 models on Novita support function calling (tool use).
Yes, Novita offers 37 reasoning models: DeepSeek V4.1 Flash 1017 (max), GLM-5.3 (max), MiniMax-M3.1-Flash, GLM-5.3-Flash, DeepSeek V4 Pro 0813 (max), MiniMax-M3, Gemma 4 31B, gpt-oss-120b (high), Kimi K2.6, Qwen3.5 397B A17B, MiniMax-M2.5, Qwen3.5 27B, Qwen3.5 122B A10B, GLM-4.7-Flash, GLM-4.5V, GLM-5 FP8, Qwen3.5 35B A3B, MiniMax-M2.7 (FP8), GLM-5.1 (FP8), GLM-4.6, Gemma 4 26B A4B, GLM-4.7, Nemotron 3 Nano (FP4), Qwen3.6 35B A3B, Qwen3.6 27B, Kimi K2.7 Code, MiniMax M1 80k, Kimi K2.5, DeepSeek R1 0528, Hy3, Kimi K2 Thinking, DeepSeek V4 Pro (max), gpt-oss-120b (low), GLM-4.6V, DeepSeek V4.1 Flash (max), GLM-5.2 (max) (FP8), and DeepSeek V4 Flash 0731 (max). Reasoning models use extended thinking to work through complex problems before providing an answer.
Yes, all 61 models on Novita are open weight models.
Yes, provider performance can vary over time due to infrastructure changes, load balancing, and updates. We continuously benchmark all providers and display historical performance trends in the "Over Time" charts.
When choosing a model on Novita, consider: intelligence (for quality-sensitive tasks), time per task (how quickly the model completes an Intelligence Index task), latency (for interactive applications requiring quick first responses), pricing (for cost-sensitive workloads), and features like context window size, JSON mode, or function calling support.