Gemma 3 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 | 99.63%◆ (best) | 99.38% | 99.27%◆ (best) | 99.27%◆ (best) | 611ms | 4.56s◆ (best) | 24 tps | 131K | $0.080 / $0.450◆ (best) | |
| Healthy | 99.39% | 76.23% | 90.55% | 90.55% | 644ms | 4.59s | 29 tps | 98K | $0.119 / $0.200 | |
| Healthy | 98.23% | 99.61% | 98.69% | 98.69% | 1.02s | 7.20s | 20 tps | 131K | $0.080 / $0.160◆ (best) | |
Phalaphala | Degraded | 91.32% | 94.70% | 95.07% | 95.07% | 474ms◆ (best) | 5.57s | 85 tps◆ (best) | 262K | $0.150 / $0.460 |
| Down | 74.69% | 78.52% | 86.67% | 86.67% | 1.25s | 12.5s | 11 tps | 110K | $0.100 / $0.300 |
- Parasail99.63%◆ (best)Healthy24h 99.38%
- 7d
- 99.27%◆ (best)
- 30d
- 99.27%◆ (best)
- p50
- 611ms
- p99
- 4.56s◆ (best)
- Throughput
- 24 tps
- Context
- 131K
- $/Mtok
- $0.080 / $0.450◆ (best)
parasail/fp8fp8 - Novita99.39%Healthy24h 76.23%
- 7d
- 90.55%
- 30d
- 90.55%
- p50
- 644ms
- p99
- 4.59s
- Throughput
- 29 tps
- Context
- 98K
- $/Mtok
- $0.119 / $0.200
novita/bf16bf16 - DeepInfra98.23%Healthy24h 99.61%
- 7d
- 98.69%
- 30d
- 98.69%
- p50
- 1.02s
- p99
- 7.20s
- Throughput
- 20 tps
- Context
- 131K
- $/Mtok
- $0.080 / $0.160◆ (best)
deepinfra/fp8fp8 - Phala91.32%Degraded24h 94.70%
- 7d
- 95.07%
- 30d
- 95.07%
- p50
- 474ms◆ (best)
- p99
- 5.57s
- Throughput
- 85 tps◆ (best)
- Context
- 262K
- $/Mtok
- $0.150 / $0.460
phala - Nebius74.69%Down24h 78.52%
- 7d
- 86.67%
- 30d
- 86.67%
- p50
- 1.25s
- p99
- 12.5s
- Throughput
- 11 tps
- Context
- 110K
- $/Mtok
- $0.100 / $0.300
nebius/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.
Throughput
Median tokens per second.