Provider reliability · tracked daily
OpenAI: gpt-oss-20b (batch): which provider is most reliable?
Across 13 providers, Phala is the most reliable way to run OpenAI: gpt-oss-20b (batch), at 100.0% uptime and 377ms median latency. Measured directly from OpenRouter's endpoint data, tracked daily.
- Phala - 100.0% uptime · 377ms
- CoreWeave - 100.0% uptime · 116ms
- Novita - 100.0% uptime · 735ms
- Parasail - 100.0% uptime · 443ms
- SiliconFlow - 99.9% uptime · 1256ms
- DeepInfra - 99.9% uptime · 247ms
- Darkbloom - 99.9% uptime · 2787ms
- Groq - 99.9% uptime · 546ms
- AkashML - 99.4% uptime · 1478ms
- Google - 79.7% uptime · 386ms
Uptime is OpenRouter's own trailing-30-minute metric, captured once at our last daily sync - a snapshot of that window, not a live/real-time feed.
| # | Provider | Uptime (30m) | Latency p50 | Throughput |
|---|---|---|---|---|
| 01 | Phala | 100.0% | 377ms | 69 tok/s |
| 02 | CoreWeave | 100.0% | 116ms | 132 tok/s |
| 03 | Novita | 100.0% | 735ms | 104 tok/s |
| 04 | Parasail | 100.0% | 443ms | 58 tok/s |
| 05 | SiliconFlow | 99.9% | 1256ms | 44 tok/s |
| 06 | DeepInfra | 99.9% | 247ms | 90 tok/s |
| 07 | Darkbloom | 99.9% | 2787ms | 33 tok/s |
| 08 | Groq | 99.9% | 546ms | 453 tok/s |
| 09 | AkashML | 99.4% | 1478ms | 25 tok/s |
| 10 | 79.7% | 386ms | 168 tok/s | |
| 11 | Fireworks | 0.0% | - | - |
| 12 | Together | - | 286ms | 61 tok/s |
| 13 | Amazon Bedrock | - | - | - |
Uptime over time
- AkashML
- Amazon Bedrock
- CoreWeave
- Darkbloom
- DeepInfra
- Fireworks
Frequently asked questions
Which provider is most reliable for OpenAI: gpt-oss-20b (batch)?
Phala is the most reliable provider for OpenAI: gpt-oss-20b (batch), at 100.0% uptime (OpenRouter's trailing-30-minute metric, snapshotted at our last daily sync) and 377ms median latency, out of 13 providers tracked.
What is OpenAI: gpt-oss-20b (batch)'s uptime?
The best-performing provider for OpenAI: gpt-oss-20b (batch) reports 100.0% uptime, tracked daily.