Qwen3.5-27B: 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 | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Healthy | 100%◆ (best) | 98.76% | 97.88% | 97.88% | 415ms◆ (best) | 10.8s | 33 tps◆ (best) | 262K | $0.260 / $2.60 | |
| Healthy | 100%◆ (best) | 99.71% | 99.80%◆ (best) | 99.80%◆ (best) | 807ms | 5.30s◆ (best) | 17 tps | 262K | $0.300 / $2.40 | |
| Healthy | 100%◆ (best) | 99.14% | 98.98% | 98.98% | 1.21s | 7.22s | 10 tps | 262K | $0.250 / $2.00 | |
| Healthy | 98.90% | 99.30% | 99.02% | 99.02% | 1.24s | 6.26s | 19 tps | 262K | $0.270 / $2.16 | |
Alibabaalibaba | Down | 74.05% | 97.37% | 98.02% | 98.02% | 908ms | 6.62s | 23 tps | 262K | $0.195 / $1.56◆ (best) |
Phalaphala | No data | — | 76.18% | 64.96% | 64.96% | 1.18s | 11.3s | 17 tps | 262K | $0.300 / $2.40 |
- DeepInfra100%◆ (best)Healthy24h 98.76%
- 7d
- 97.88%
- 30d
- 97.88%
- p50
- 415ms◆ (best)
- p99
- 10.8s
- Throughput
- 33 tps◆ (best)
- Context
- 262K
- $/Mtok
- $0.260 / $2.60
deepinfra/fp8fp8 - Novita100%◆ (best)Healthy24h 99.71%
- 7d
- 99.80%◆ (best)
- 30d
- 99.80%◆ (best)
- p50
- 807ms
- p99
- 5.30s◆ (best)
- Throughput
- 17 tps
- Context
- 262K
- $/Mtok
- $0.300 / $2.40
novita/bf16bf16 - SiliconFlow100%◆ (best)Healthy24h 99.14%
- 7d
- 98.98%
- 30d
- 98.98%
- p50
- 1.21s
- p99
- 7.22s
- Throughput
- 10 tps
- Context
- 262K
- $/Mtok
- $0.250 / $2.00
siliconflow/fp8fp8 - AtlasCloud98.90%Healthy24h 99.30%
- 7d
- 99.02%
- 30d
- 99.02%
- p50
- 1.24s
- p99
- 6.26s
- Throughput
- 19 tps
- Context
- 262K
- $/Mtok
- $0.270 / $2.16
atlas-cloud/fp8fp8 - Alibaba74.05%Down24h 97.37%
- 7d
- 98.02%
- 30d
- 98.02%
- p50
- 908ms
- p99
- 6.62s
- Throughput
- 23 tps
- Context
- 262K
- $/Mtok
- $0.195 / $1.56◆ (best)
alibaba - No data24h 76.18%
- 7d
- 64.96%
- 30d
- 64.96%
- p50
- 1.18s
- p99
- 11.3s
- Throughput
- 17 tps
- Context
- 262K
- $/Mtok
- $0.300 / $2.40
phala
Measured 55s 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.
Throughput
Median tokens per second.