GMI: Models Intelligence, Performance & Price
Analysis of GMI'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 GMI for your use-case.
Most Intelligent
Intelligence index
Total 6 models
Fastest
Output speed
Total 6 models
Lowest Price
Blended price (per 1M tokens)
Total 6 models
GMI offers 6 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 GMI are Kimi K2.7 Code (FP8) (42), Hy3 (41), and MiniMax-M2.5 FP8 (34).
- For output speed, the fastest models are Qwen3 Next 80B A3B (181 t/s), MiniMax-M2.5 FP8 (89 t/s), and Hy3 (69 t/s). Speed varies significantly across models, with a 340% difference between the fastest and slowest.
- For latency, Qwen3 Next 80B A3B (2.36s), Kimi K2.5 (3.92s), and DeepSeek V3.2 (4.00s) offer the lowest time to first answer token.
- For pricing, Hy3 ($0.10), Qwen3 Next 80B A3B ($0.29), and DeepSeek V3.2 ($0.30) offer the lowest blended prices per 1M tokens. Prices vary up to 7.1x across models.
- For context window size, Hy3 (262k), Kimi K2.5 (262k), and Qwen3 Next 80B A3B (262k) support the largest context windows on GMI.
Highlights
Intelligence Evaluations
Artificial Analysis Intelligence Index
Intelligence Evaluations
Agentic real-world work tasks, (Elo-500)/2000
Agentic tool use
Agentic coding & terminal use
Coding
Reasoning & knowledge
Scientific reasoning
Physics reasoning
Knowledge
1 - hallucination rate
Long context reasoning
Agentic knowledge work, Elo
Agentic SaaS workflows
Legal agentic work, criterion pass rate
Agentic business operations
Instruction following
Long-horizon agentic tasks
Kubernetes incident root-cause analysis
Visual reasoning
Intelligence Index vs. Price
Context Window
Context Window
Pricing
Intelligence Index vs. Price
Performance Summary
Output Speed vs. Price
Speed
Measured by Output Speed (tokens per second)
Output Speed
Latency
Measured by Time (seconds) to First Token
Latency: Time To First Answer Token
End-to-End Response Time
Seconds to output 500 tokens, calculated based on time to first token, 'thinking' time for reasoning models, and output speed
End-to-End Response Time vs. Price
Further Analysis | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
 GLM-5.2 (max) (FP8) | 1.05M | Open | 51 | -- | -- | -- | -- | -- | |||
MiniMax-M3 | 1.05M | Open | 44 | -- | -- | -- | -- | -- | |||
DeepSeek V4 Pro (max) | 1.05M | Open | 44 | -- | -- | -- | -- | -- | |||
Kimi K2.6 FP8 | 262k | Open | 44 | -- | -- | -- | -- | -- | |||
 MiMo-V2.5-Pro | 1.05M | Open | 42 | -- | -- | -- | -- | -- | |||
Kimi K2.7 Code (FP8) | 65.5k | Open | 42 | $0.46 | 44 | 6.05 | 67.84 | 50.46 | |||
Hy3 | 262k | Open | 41 | $0.03 | 69 | 3.69 | 40.07 | 29.10 | |||
DeepSeek V4 Flash (max) | 1.05M | Open | 40 | -- | -- | -- | -- | -- | |||
GLM-5.1 (FP8) | 203k | Open | 40 | -- | -- | -- | -- | -- | |||
MiniMax-M2.7 (FP8) | 197k | Open | 38 | -- | -- | -- | -- | -- | |||
 Nemotron 3 Ultra | 262k | Open | 38 | -- | -- | -- | -- | -- | |||
DeepSeek V4 Flash (high) | 1.05M | Open | 37 | -- | -- | -- | -- | -- | |||
MiMo-V2.5 | 1.05M | Open | 37 | $0.02 | -- | -- | -- | -- | |||
Qwen3.5 27B (FP8) | 262k | Open | 34* | -- | -- | -- | -- | -- | |||
Qwen3.5 397B A17B (FP8) | 262k | Open | 34 | $0.36 | -- | -- | -- | -- | |||
MiniMax-M2.5 FP8 | 197k | Open | 34* | -- | 89 | 2.06 | 30.03 | 22.38 | |||
Hy3-preview | 262k | Open | 34* | -- | -- | -- | -- | -- | |||
Qwen3.5 122B A10B (FP8) | 262k | Open | 32 | $0.26 | -- | -- | -- | -- | |||
Qwen3.6 35B A3B FP8 | 262k | Open | 32 | $0.19 | -- | -- | -- | -- | |||
Kimi K2.5 | 262k | Open | 29* | -- | 38 | 3.92 | 17.14 | -- | |||
 Gemma 4 31B (FP8) | 262k | Open | 29 | $0.04 | -- | -- | -- | -- | |||
Qwen3.5 35B A3B (FP8) | 262k | Open | 29* | -- | -- | -- | -- | -- | |||
DeepSeek V4 Flash | 1.05M | Open | 29* | -- | -- | -- | -- | -- | |||
MiMo-V2.5-Pro | 1.05M | Open | 28* | -- | -- | -- | -- | -- | |||
Hy3-preview | 262k | Open | 26* | -- | -- | -- | -- | -- | |||
Gemma 4 26B A4B (FP8) | 1.05M | Open | 26 | $0.04 | -- | -- | -- | -- | |||
DeepSeek V3.2 | 164k | Open | 25* | -- | 41 | 4.00 | 16.17 | -- | |||
Qwen3.6 35B A3B FP8 | 262k | Open | 24 | $0.42 | -- | -- | -- | -- | |||
Gemma 4 26B A4B (FP8) | 1.05M | Open | 20* | -- | -- | -- | -- | -- | |||
Qwen3 Next 80B A3B | 262k | Open | 17 | $0.05 | -- | -- | -- | -- | |||
Qwen3 Next 80B A3B | 262k | Open | 14* | -- | 181 | 2.36 | 5.13 | -- | |||
Key definitions
Frequently Asked Questions
Common questions about GMI
GMI offers 6 models that we track: Kimi K2.7 Code (FP8), Hy3, MiniMax-M2.5 FP8, Kimi K2.5, DeepSeek V3.2, and Qwen3 Next 80B A3B.
The most intelligent model available on GMI is Kimi K2.7 Code (FP8) with an Intelligence Index score of 42.
The fastest model on GMI by output speed is Qwen3 Next 80B A3B at 180.7 tokens per second.
The model with the lowest time to first answer token on GMI is Qwen3 Next 80B A3B at 2.36s. Lower latency means faster initial response time.
The most affordable model on GMI by blended price is Hy3 at $0.10 per 1M tokens (7:2:1 cache hit/input/output ratio).
Yes, GMI offers an OpenAI-compatible API, making it easy to switch from OpenAI or use existing OpenAI SDK integrations.
Yes, all 6 models on GMI support JSON mode for structured output.
Yes, all 6 models on GMI support function calling (tool use).
Yes, GMI offers 3 reasoning models: Kimi K2.7 Code (FP8), Hy3, and MiniMax-M2.5 FP8. Reasoning models use extended thinking to work through complex problems before providing an answer.
Yes, all 6 models on GMI 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 GMI, consider: intelligence (for quality-sensitive tasks), output speed (for throughput-intensive tasks), 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.