Provides OpenAI-compatible API access to open-source AI models like DeepSeek and QwQ, powered by shared idle computing resources.
Performance Stats
Avg Speed
25.15t/s
Latency
25.34s
Total Tests
13,724
Models
15
| Model | Speed | Latency | Tests |
|---|---|---|---|
| best-free | 101.41 t/s | 17.62s | 10 |
| deepseek-r1:7b | 47.22 t/s | 5.66s | 385 |
| Qwen/QwQ-32B | 41.50 t/s | 25.84s | 15 |
| Qwen/Qwen3-VL-32B-Thinking | 36.49 t/s | 20.51s | 40 |
| Qwen/Qwen3-VL-32B-Thinking | 36.49 t/s | 20.51s | 40 |
| zhipu/glm-4-flash | 35.00 t/s | 1.78s | 10 |
| QwQ-32B | 28.31 t/s | 2.66s | 45 |
| free:QwQ-32B | 28.02 t/s | 26.45s | 6585 |
| free:Qwen3-30B-A3B | 20.87 t/s | 25.78s | 6585 |
| zai-org/GLM-4.5 | 20.72 t/s | 1.47s | 5 |
| zai-org/GLM-4.5 | 20.72 t/s | 1.47s | 5 |
| 84cf9645 | 0.00 t/s | 0.00s | 15 |
| 你好 | 0.00 t/s | 0.00s | 10 |
| Spark Lite | 0.00 t/s | 0.00s | 10 |
| wen3-30B-A3B | 0.00 t/s | 0.00s | 5 |
| Time | Model | Speed | Latency |
|---|---|---|---|
| Jan 15, 01:36 PM | Qwen/Qwen3-VL-32B-Thinking | 44.46 t/s | 14.29s |
| Jan 14, 09:05 AM | free:QwQ-32B | 0.00 t/s | 0.00s |
| Jan 14, 09:04 AM | free:QwQ-32B | 0.00 t/s | 0.00s |
| Jan 13, 02:52 PM | Qwen/QwQ-32B | 46.81 t/s | 18.74s |
| Jan 9, 05:26 PM | free:Qwen3-30B-A3B | 17.28 t/s | 6.02s |
| Jan 9, 02:06 AM | free:Qwen3-30B-A3B | 16.13 t/s | 8.91s |
| Jan 8, 07:25 AM | free:Qwen3-30B-A3B | 16.11 t/s | 21.90s |
| Jan 3, 08:33 PM | zai-org/GLM-4.5 | 20.72 t/s | 1.47s |
| Jan 3, 01:56 PM | free:Qwen3-30B-A3B | 20.64 t/s | 5.72s |
| Dec 30, 02:55 PM | Qwen/Qwen3-VL-32B-Thinking | 38.56 t/s | 17.49s |