GMI: Models Intelligence, Performance & Price
This analysis is intended to support you in choosing the best model provided by GMI for your use-case.
Most Intelligent
Intelligence Index
Total 17 models
Lowest Time per Task
Time per Intelligence Index task (minutes)
Total 17 models
Lowest Price
Blended price (per 1M tokens)
Total 17 models
GMI offers 17 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 DeepSeek V4.1 Flash 1017 (max) (33), GLM-5.3 (max) (32), and MiMo-V2.6-Pro (BF16) (30).
- For time per task, DeepSeek V4.1 Flash 1017 (max) (11.3 min), MiniMax-M3.1-Flash (12.1 min), and MiniMax-M3 (15.6 min) finish an Intelligence Index task fastest on GMI.
- For latency, MiniMax-M3 (1.55s), Qwen3.6 35B A3B (non-reasoning) FP8 (2.95s), and Kimi K2.5 (non-reasoning) (9.48s) offer the lowest time to first answer token.
- For pricing, MiMo-V2.6-Flash (BF16) ($0.05), GLM-5.3-Flash ($0.05), and MiniMax-M3.1-Flash ($0.08) offer the lowest blended prices per 1M tokens. Prices vary up to 2.4x across models.
- For context window size, DeepSeek V4 Pro 0813 (max) (1M), MiniMax-M3 (1M), and DeepSeek V4.1 Flash 1017 (max) (1M) support the largest context windows on GMI.
- DeepSeek V4.1 Flash 1017 (max) offers the best combination of intelligence and time per task. For cost optimization, MiMo-V2.6-Flash (BF16) provides the most competitive pricing.
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 | $0.30 | 13.0 | 41.46 | 16.4% | 1M | Open | |||
GLM-5.3 (max) | 32 | $2.53 | 26.5 | 68.15 | 1.7% | 1M | Open | |||
MiMo-V2.6-Pro (BF16) | 30 | $0.19 | 58.7 | 42.15 | 23.5% | 1M | Open | |||
MiniMax-M3.1-Flash | 29 | $0.33 | 14.0 | 27.59 | 3.8% | 1M | Open | |||
GLM-5.3-Flash | 28 | $0.41 | 48.6 | 84.77 | 4.0% | 1M | Open | |||
MiMo-V2.6-Flash (BF16) | 27 | $0.11 | 66.3 | 30.94 | 2.0% | 524k | Open | |||
DeepSeek V4 Pro 0813 (max) | 23 | $2.66 | 22.9 | 60.30 | 0.0% | 1M | Open | |||
MiniMax-M3 | 19 | $0.91 | 15.0 | 1.71 | 0.0% | 1M | Open | |||
MiMo-V2.5-Pro | -- | -- | -- | -- | -- | 1M | Open | |||
Kimi K2.6 FP8 | -- | -- | -- | -- | -- | 262k | Open | |||
MiniMax-M2.5 FP8 | -- | -- | -- | 27.15 | 0.0% | 197k | Open | |||
Qwen3.5 27B (FP8) | -- | -- | -- | -- | -- | 262k | Open | |||
MiMo-V2.5 | -- | -- | -- | -- | -- | 1M | Open | |||
Qwen3.5 35B A3B (FP8) | -- | -- | -- | -- | -- | 262k | Open | |||
MiMo-V2.5-Pro (non-reasoning) | -- | -- | -- | -- | -- | 1M | Open | |||
MiniMax-M2.7 (FP8) | -- | -- | -- | -- | -- | 197k | Open | |||
GLM-5.1 (FP8) | -- | -- | -- | -- | -- | 203k | Open | |||
Qwen3.6 35B A3B (non-reasoning) FP8 | -- | -- | -- | 2.91 | 0.0% | 262k | Open | |||
DeepSeek V3.2 (non-reasoning) | -- | -- | -- | -- | -- | 164k | Open | |||
DeepSeek V4 Flash Vision (max) | -- | -- | -- | 50.02 | 0.0% | 1M | Proprietary | |||
Qwen3.6 35B A3B FP8 | -- | -- | -- | 29.25 | 0.8% | 262k | Open | |||
DeepSeek V4 Flash (high) | -- | -- | -- | -- | -- | 1M | Open | |||
Hy3-preview | -- | -- | -- | -- | -- | 262k | Open | |||
Kimi K2.7 Code (FP8) | -- | -- | -- | 21.42 | 1.5% | 66k | Open | |||
DeepSeek V4 Flash (non-reasoning) | -- | -- | -- | -- | -- | 1M | Open | |||
Hy3 | -- | -- | -- | 46.35 | 0.0% | 262k | Open | |||
Hy3-preview (non-reasoning) | -- | -- | -- | -- | -- | 262k | Open | |||
Kimi K2.5 (non-reasoning) | -- | -- | -- | 10.38 | 21.7% | 262k | Open | |||
DeepSeek V4 Pro (max) | -- | -- | -- | -- | -- | 1M | Open | |||
DeepSeek V4 Flash (max) | -- | -- | -- | -- | -- | 1M | Open | |||
DeepSeek V4.1 Flash (max) | -- | -- | -- | 54.78 | 0.0% | 1M | Open | |||
GLM-5.2 (max) (FP8) | -- | -- | -- | -- | -- | 1M | Open | |||
DeepSeek V4 Flash 0731 (max) | -- | -- | -- | 38.18 | 0.0% | 1M | Open | |||
Key definitions
Frequently Asked Questions
Common questions about GMI
GMI offers 17 models that we track: DeepSeek V4.1 Flash 1017 (max), GLM-5.3 (max), MiMo-V2.6-Pro (BF16), MiniMax-M3.1-Flash, GLM-5.3-Flash, MiMo-V2.6-Flash (BF16), DeepSeek V4 Pro 0813 (max), MiniMax-M3, MiniMax-M2.5 FP8, Qwen3.6 35B A3B (non-reasoning) FP8, DeepSeek V4 Flash Vision (max), Qwen3.6 35B A3B FP8, Kimi K2.7 Code (FP8), Hy3, Kimi K2.5 (non-reasoning), DeepSeek V4.1 Flash (max), and DeepSeek V4 Flash 0731 (max).
The most intelligent model available on GMI is DeepSeek V4.1 Flash 1017 (max) with an Intelligence Index score of 33.
The model with the lowest time per task on GMI is DeepSeek V4.1 Flash 1017 (max) at 11.3 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 GMI is MiniMax-M3 at 1.55s. Lower latency means faster initial response time.
The most affordable model on GMI by blended price is MiMo-V2.6-Flash (BF16) at $0.0468 per 1M tokens (7:2:1 cache hit/input/output ratio).
Prices on GMI vary up to 17.9x across models, from $0.0468 per 1M tokens for MiMo-V2.6-Flash (BF16) to $0.84 per 1M tokens for Kimi K2.5 (non-reasoning).
Yes, GMI offers an OpenAI-compatible API, making it easy to switch from OpenAI or use existing OpenAI SDK integrations.
Yes, all 17 models on GMI support JSON mode for structured output.
Yes, all 17 models on GMI support function calling (tool use).
Yes, GMI offers 15 reasoning models: DeepSeek V4.1 Flash 1017 (max), GLM-5.3 (max), MiMo-V2.6-Pro (BF16), MiniMax-M3.1-Flash, GLM-5.3-Flash, MiMo-V2.6-Flash (BF16), DeepSeek V4 Pro 0813 (max), MiniMax-M3, MiniMax-M2.5 FP8, DeepSeek V4 Flash Vision (max), Qwen3.6 35B A3B FP8, Kimi K2.7 Code (FP8), Hy3, DeepSeek V4.1 Flash (max), and DeepSeek V4 Flash 0731 (max). Reasoning models use extended thinking to work through complex problems before providing an answer.
Yes, 16 of 17 models on GMI are open weight models: DeepSeek V4.1 Flash 1017 (max), GLM-5.3 (max), MiMo-V2.6-Pro (BF16), MiniMax-M3.1-Flash, GLM-5.3-Flash, MiMo-V2.6-Flash (BF16), DeepSeek V4 Pro 0813 (max), MiniMax-M3, MiniMax-M2.5 FP8, Qwen3.6 35B A3B (non-reasoning) FP8, Qwen3.6 35B A3B FP8, Kimi K2.7 Code (FP8), Hy3, Kimi K2.5 (non-reasoning), DeepSeek V4.1 Flash (max), and DeepSeek V4 Flash 0731 (max).
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), 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.