Hy4 preview: 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) | 95.21% | 99.11% | 98.96% | 1.56s◆ (best) | 2.85s◆ (best) | 59 tps◆ (best) | 1.0M | $0.834 / $2.50 | |
| Healthy | 100%◆ (best) | 96.92% | 97.01% | 98.98% | 3.83s | 13.5s | 48 tps | 1.0M | $0.834 / $2.50 | |
| No data | — | 98.10% | 99.75%◆ (best) | 99.75%◆ (best) | 7.60s | 10.4s | 13 tps | 1.0M | $0.834 / $2.50 | |
| No data | — | — | 95.81% | 95.81% | — | — | — | 1.0M | $0.747 / $2.25◆ (best) | |
| No data | — | 80.86% | 95.08% | 96.94% | 3.88s | 18.6s | 32 tps | 1.0M | $0.834 / $2.50 |
- SiliconFlow100%◆ (best)Healthy24h 95.21%
- 7d
- 99.11%
- 30d
- 98.96%
- p50
- 1.56s◆ (best)
- p99
- 2.85s◆ (best)
- Throughput
- 59 tps◆ (best)
- Context
- 1.0M
- $/Mtok
- $0.834 / $2.50
siliconflow/fp8fp8 - Tencent100%◆ (best)Healthy24h 96.92%
- 7d
- 97.01%
- 30d
- 98.98%
- p50
- 3.83s
- p99
- 13.5s
- Throughput
- 48 tps
- Context
- 1.0M
- $/Mtok
- $0.834 / $2.50
tencent/fp8fp8 - No data24h 98.10%
- 7d
- 99.75%◆ (best)
- 30d
- 99.75%◆ (best)
- p50
- 7.60s
- p99
- 10.4s
- Throughput
- 13 tps
- Context
- 1.0M
- $/Mtok
- $0.834 / $2.50
deepinfra/fp8fp8 - No data24h —
- 7d
- 95.81%
- 30d
- 95.81%
- p50
- —
- p99
- —
- Throughput
- —
- Context
- 1.0M
- $/Mtok
- $0.747 / $2.25◆ (best)
inference-net/fp8fp8 - No data24h 80.86%
- 7d
- 95.08%
- 30d
- 96.94%
- p50
- 3.88s
- p99
- 18.6s
- Throughput
- 32 tps
- Context
- 1.0M
- $/Mtok
- $0.834 / $2.50
novita/fp8fp8
Measured 4m 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.