Fast AI models: low latency
Sometimes the metric is milliseconds, not the leaderboard. Autocomplete, hot-path classification, and UIs that cannot wait. This collection prioritizes flash/lite/mini/highspeed variants.
Models in this collection
| Model | Type | Context | Price |
|---|---|---|---|
 Gemini 2.5 Flash-LiteGoogle | Chat | 1M tokens | $0.109/M in · $0.436/M out |
| Gemini 3.5 Flash-LiteGoogle | Chat | 1M tokens | $0.327/M in · $2.7251/M out |
| Gemini 3.1 Flash-LiteGoogle | Chat | 1M tokens | $0.2725/M in · $1.6351/M out |
 DeepSeek V4 FlashDeepSeek | Chat | 1M tokens | $0.1526/M in · $0.3052/M out |
 GPT-4.1 mini (legacy)OpenAI | Chat | 1M tokens | $0.436/M in · $1.7441/M out |
 Claude Haiku 4.5Anthropic | Chat | 200k tokens | $1.0901/M in · $5.4503/M out |
 Ministral 8BMistral | Chat | 128k tokens | $0.109/M in · $0.109/M out |
 Kimi K2.7 Code · HighspeedMoonshot (Kimi) | Chat | 262k tokens | $1.0356/M in · $8.7205/M out |
| Gemini 2.5 FlashGoogle | Chat | 1M tokens | $0.327/M in · $2.7251/M out |
Why these models
Flash Lite and DeepSeek Flash for the latency floor; Haiku and GPT mini when you want speed with more judgment; Kimi Highspeed for concurrent code chat.
Use them with Geek Hub
One OpenAI-compatible base URL and API key. Swap any model id from this list without rewriting your client.
Get an API keyFAQ
- Is fast the same as cheap?
- They often correlate, not always. Check price and p95 latency in your region with a smoke test.
- Good for agents?
- As workers, yes. As the only planner, sometimes no. Typical pattern: fast router + strong planner.
- Does streaming help?
- Yes for perceived speed. Enable stream on the client even when the model is already fast.