Qwen3 30B A3B Instruct 2507: providers compared
Every endpoint serving this model, side by side. ◆ marks the best value in each column — per model, never fleet-wide.
Current state
| State | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
Alibabaalibaba | Healthy | 100%◆ (best) | 99.95% | 99.96%◆ (best) | 99.96%◆ (best) | 431ms | 2.40s | 37 tps | 131K | $0.130 / $0.520 |
| Healthy | 100%◆ (best) | 98.54% | 97.12% | 97.12% | 344ms◆ (best) | 1.57s◆ (best) | 70 tps◆ (best) | 262K | $0.100 / $0.300 | |
StreamLakestreamlake | Healthy | 99.83% | 99.88% | 99.86% | 99.86% | 1.58s | 8.09s | 17 tps | 128K | $0.048 / $0.193◆ (best) |
| Healthy | 98.67% | 98.27% | 96.91% | 96.91% | 360ms | 6.57s | 30 tps | 262K | $0.100 / $0.300 | |
| Healthy | 98.39% | 90.36% | 90.45% | 90.45% | 2.01s | 13.3s | 14 tps | 262K | $0.090 / $0.300 |
- Alibaba100%◆ (best)Healthy24h 99.95%
- 7d
- 99.96%◆ (best)
- 30d
- 99.96%◆ (best)
- p50
- 431ms
- p99
- 2.40s
- Throughput
- 37 tps
- Context
- 131K
- $/Mtok
- $0.130 / $0.520
alibaba - CoreWeave100%◆ (best)Healthy24h 98.54%
- 7d
- 97.12%
- 30d
- 97.12%
- p50
- 344ms◆ (best)
- p99
- 1.57s◆ (best)
- Throughput
- 70 tps◆ (best)
- Context
- 262K
- $/Mtok
- $0.100 / $0.300
coreweave/bf16bf16 - StreamLake99.83%Healthy24h 99.88%
- 7d
- 99.86%
- 30d
- 99.86%
- p50
- 1.58s
- p99
- 8.09s
- Throughput
- 17 tps
- Context
- 128K
- $/Mtok
- $0.048 / $0.193◆ (best)
streamlake - Nebius98.67%Healthy24h 98.27%
- 7d
- 96.91%
- 30d
- 96.91%
- p50
- 360ms
- p99
- 6.57s
- Throughput
- 30 tps
- Context
- 262K
- $/Mtok
- $0.100 / $0.300
nebius/fp8fp8 - SiliconFlow98.39%Healthy24h 90.36%
- 7d
- 90.45%
- 30d
- 90.45%
- p50
- 2.01s
- p99
- 13.3s
- Throughput
- 14 tps
- Context
- 262K
- $/Mtok
- $0.090 / $0.300
siliconflow/fp8fp8
Measured 2m ago. Latency and throughput are OpenRouter's rolling 30-minute windows. ◆ marks the best value in each column among these endpoints.
Price vs latency
One dot per endpoint at the latest round · further left = faster, lower = cheaper · hover a dot for its tag
History
Availability
Axis starts at 90% so small differences stay visible; it expands to 0 when an endpoint falls below.
Time to first token
Logarithmic axis — the fleet spans two orders of magnitude. P50 is OpenRouter's rolling 30-minute window, averaged over the hour.
Throughput
Median tokens per second.