Stability ranking last 24 hours
A stability score out of 100 that combines success rate, latency and latency variation over the last 24 hours; higher is better. Success rate is taken conservatively by sample size, so a route with few samples cannot rank high just because every request happened to succeed.
Data29 routes · 7 relaysSynced 11/13/2026, 05:09:34, updated every 15 minutes
Model: Claude Fable 5.1GPT 6 AstraGPT 6.1 SolClaude Haiku 5.5GPT 6 SolClaude Opus 5.5DeepSeek V4.1 FlashGrok 4.7
| Rank | Model @ relay | Stability | Samples | Availability | Session cost | Data security |
|---|---|---|---|---|---|---|
| 1 | 95.7 | 170 | 98.2% | ¥4.38 | 4 | |
| 2 | 95.5 | 135 | 98.5% | ¥2.19 | 4 | |
| 3 | 95.4 | 148 | 98.7% | ¥0.44 | 4 | |
| 4 | 95.4 | 200 | 98.0% | — | 4 | |
| 5 | 94.6 | 132 | 97.7% | ¥0.44 | 4 | |
| 6 | 94.5 | 146 | 97.3% | ¥1.75 | 4 | |
| 7 | 86.8 | 217 | 98.2% | ¥2.59 | 4 | |
| 8 | 86.6 | 176 | 98.3% | ¥3.10 | 4 | |
| 9 | 75.1 | 31 | 100.0% | — | 4 | |
| 10 | 74.3 | 31 | 100.0% | — | 4 | |
| 11 | 74.3 | 30 | 96.7% | — | 4 | |
| 12 | 66.8 | 30 | 93.3% | — | 4 | |
| 13 | 66.5 | 30 | 73.3% | — | 4 | |
| 14 | 48.2 | 31 | 77.4% | — | 3 | |
| 15 | 82.3 | 31 | 100.0% | — | 0 | |
| 16 | 82.0 | 30 | 100.0% | — | 0 | |
| 17 | 82.0 | 30 | 100.0% | ¥2.38 | 0 | |
| 18 | 80.2 | 30 | 100.0% | — | 0 | |
| 19 | 78.3 | 30 | 96.7% | — | 0 | |
| 20 | 75.5 | 30 | 100.0% | — | 0 | |
| 21 | 73.3 | 29 | 100.0% | — | 0 | |
| 22 | 71.7 | 31 | 100.0% | — | 0 | |
| 23 | 70.5 | 30 | 93.3% | ¥11.9 | 0 | |
| 24 | 70.4 | 29 | 89.7% | — | 0 | |
| 25 | 69.2 | 30 | 86.7% | — | 0 | |
| 26 | 68.6 | 30 | 90.0% | — | 0 | |
| 27 | 67.2 | 30 | 96.7% | — | 0 | |
| 28 | 48.5 | 29 | 79.3% | — | 0 | |
| 29 | 43.4 | 30 | 60.0% | ¥4.77 | 0 |
FAQ
How is stability calculated?
A stability score out of 100 that combines success rate, latency and latency variation over the last 24 hours; higher is better. Success rate is taken conservatively by sample size, so a route with few samples cannot rank high just because every request happened to succeed.
Why does the same relay appear more than once?
Each row is one “model @ relay” route. Different models and groups at the same relay can perform very differently, so they are ranked separately.