gpt-oss-120b: 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) | 99.63% | 99.17% | 99.17% | 254ms | 812ms◆ (best) | 195 tps | 128K | $0.100 / $0.500 | |
| Healthy | 100%◆ (best) | 99.99% | 99.99% | 99.99% | 241ms◆ (best) | 2.33s | 791 tps◆ (best) | 131K | $0.350 / $0.750 | |
| Healthy | 100%◆ (best) | 99.69% | 99.00% | 99.00% | 437ms | 1.66s | 103 tps | 131K | $0.150 / $0.600 | |
| Healthy | 100%◆ (best) | 99.69% | 99.83% | 99.83% | 345ms | 1.86s | 101 tps | 131K | $0.100 / $0.750 | |
Phalaphala | Healthy | 100%◆ (best) | 97.40% | 97.63% | 97.63% | 773ms | 2.99s | 73 tps | 131K | $0.150 / $0.600 |
Groqgroq | Healthy | 99.99% | 99.94% | 99.93% | 99.93% | 267ms | 2.38s | 319 tps | 131K | $0.150 / $0.600 |
DigitalOceandigitalocean | Healthy | 99.97% | 99.90% | 99.90% | 99.90% | 790ms | 4.19s | 21 tps | 128K | $0.055 / $0.385 |
| Healthy | 99.95% | 99.18% | 99.59% | 99.59% | 1.19s | 3.47s | 57 tps | 131K | $0.037 / $0.490 | |
Amazon Bedrockamazon-bedrock | Healthy | 99.74% | 93.00% | 99.02% | 99.02% | 405ms | 1.32s | 219 tps | 131K | $0.150 / $0.600 |
| Healthy | 99.55% | 95.30% | 96.18% | 96.18% | 415ms | 10.6s | 31 tps | 131K | $0.030 / $0.170◆ (best) | |
| Healthy | 99.51% | 99.29% | 98.79% | 98.79% | 368ms | 8.28s | 156 tps | 131K | $0.150 / $0.600 | |
| Healthy | 98.65% | 99.25% | 99.75% | 99.75% | 863ms | 7.73s | 26 tps | 131K | $0.085 / $0.500 | |
Maramara | Healthy | 98.64% | 97.46% | 95.46% | 95.46% | 2.58s | 10.8s | 126 tps | 131K | $0.150 / $0.750 |
SambaNovasambanova | Down | 97.30% | 88.58% | 86.18% | 86.18% | 863ms | 42.0s | 330 tps | 131K | $0.140 / $0.950 |
Googlegoogle-vertex/global | Degraded | 94.21% | 89.78% | 89.85% | 89.85% | 675ms | 9.40s | 62 tps | 131K | $0.090 / $0.360 |
Togethertogether | Down | 87.21% | 95.19% | 97.44% | 97.44% | 475ms | 7.46s | 37 tps | 131K | $0.150 / $0.600 |
| Down | 85.75% | 88.46% | 93.27% | 93.27% | 634ms | 28.8s | 53 tps | 131K | $0.050 / $0.250 | |
| Down | 69.55% | 96.07% | 98.02% | 98.02% | 9.41s | 12.8s | 19 tps | 131K | $0.037 / $0.170 | |
Amazon Bedrockamazon-bedrock/eu-west-1 | No data | — | — | 100%◆ (best) | 100%◆ (best) | — | — | — | 131K | $0.150 / $0.600 |
| Down | — | 82.15% | 92.66% | 92.66% | 2.68s | 17.6s | 7.0 tps | 131K | $0.050 / $0.450 |
- BaseTen100%◆ (best)Healthy24h 99.63%
- 7d
- 99.17%
- 30d
- 99.17%
- p50
- 254ms
- p99
- 812ms◆ (best)
- Throughput
- 195 tps
- Context
- 128K
- $/Mtok
- $0.100 / $0.500
baseten/fp4fp4 - Cerebras100%◆ (best)Healthy24h 99.99%
- 7d
- 99.99%
- 30d
- 99.99%
- p50
- 241ms◆ (best)
- p99
- 2.33s
- Throughput
- 791 tps◆ (best)
- Context
- 131K
- $/Mtok
- $0.350 / $0.750
cerebras/fp16fp16 - DeepInfra100%◆ (best)Healthy24h 99.69%
- 7d
- 99.00%
- 30d
- 99.00%
- p50
- 437ms
- p99
- 1.66s
- Throughput
- 103 tps
- Context
- 131K
- $/Mtok
- $0.150 / $0.600
deepinfra/turbobf16 - Parasail100%◆ (best)Healthy24h 99.69%
- 7d
- 99.83%
- 30d
- 99.83%
- p50
- 345ms
- p99
- 1.86s
- Throughput
- 101 tps
- Context
- 131K
- $/Mtok
- $0.100 / $0.750
parasail/fp4fp4 - Phala100%◆ (best)Healthy24h 97.40%
- 7d
- 97.63%
- 30d
- 97.63%
- p50
- 773ms
- p99
- 2.99s
- Throughput
- 73 tps
- Context
- 131K
- $/Mtok
- $0.150 / $0.600
phala - Groq99.99%Healthy24h 99.94%
- 7d
- 99.93%
- 30d
- 99.93%
- p50
- 267ms
- p99
- 2.38s
- Throughput
- 319 tps
- Context
- 131K
- $/Mtok
- $0.150 / $0.600
groq - DigitalOcean99.97%Healthy24h 99.90%
- 7d
- 99.90%
- 30d
- 99.90%
- p50
- 790ms
- p99
- 4.19s
- Throughput
- 21 tps
- Context
- 128K
- $/Mtok
- $0.055 / $0.385
digitalocean - AkashML99.95%Healthy24h 99.18%
- 7d
- 99.59%
- 30d
- 99.59%
- p50
- 1.19s
- p99
- 3.47s
- Throughput
- 57 tps
- Context
- 131K
- $/Mtok
- $0.037 / $0.490
akashml/bf16bf16 - Amazon Bedrock99.74%Healthy24h 93.00%
- 7d
- 99.02%
- 30d
- 99.02%
- p50
- 405ms
- p99
- 1.32s
- Throughput
- 219 tps
- Context
- 131K
- $/Mtok
- $0.150 / $0.600
amazon-bedrock - CoreWeave99.55%Healthy24h 95.30%
- 7d
- 96.18%
- 30d
- 96.18%
- p50
- 415ms
- p99
- 10.6s
- Throughput
- 31 tps
- Context
- 131K
- $/Mtok
- $0.030 / $0.170◆ (best)
coreweave/fp4fp4 - Nebius99.51%Healthy24h 99.29%
- 7d
- 98.79%
- 30d
- 98.79%
- p50
- 368ms
- p99
- 8.28s
- Throughput
- 156 tps
- Context
- 131K
- $/Mtok
- $0.150 / $0.600
nebius/fp4fp4 - Mancer 298.65%Healthy24h 99.25%
- 7d
- 99.75%
- 30d
- 99.75%
- p50
- 863ms
- p99
- 7.73s
- Throughput
- 26 tps
- Context
- 131K
- $/Mtok
- $0.085 / $0.500
mancer/fp8fp8 - Mara98.64%Healthy24h 97.46%
- 7d
- 95.46%
- 30d
- 95.46%
- p50
- 2.58s
- p99
- 10.8s
- Throughput
- 126 tps
- Context
- 131K
- $/Mtok
- $0.150 / $0.750
mara - SambaNova97.30%Down24h 88.58%
- 7d
- 86.18%
- 30d
- 86.18%
- p50
- 863ms
- p99
- 42.0s
- Throughput
- 330 tps
- Context
- 131K
- $/Mtok
- $0.140 / $0.950
sambanova - Google94.21%Degraded24h 89.78%
- 7d
- 89.85%
- 30d
- 89.85%
- p50
- 675ms
- p99
- 9.40s
- Throughput
- 62 tps
- Context
- 131K
- $/Mtok
- $0.090 / $0.360
google-vertex/global - Together87.21%Down24h 95.19%
- 7d
- 97.44%
- 30d
- 97.44%
- p50
- 475ms
- p99
- 7.46s
- Throughput
- 37 tps
- Context
- 131K
- $/Mtok
- $0.150 / $0.600
together - Novita85.75%Down24h 88.46%
- 7d
- 93.27%
- 30d
- 93.27%
- p50
- 634ms
- p99
- 28.8s
- Throughput
- 53 tps
- Context
- 131K
- $/Mtok
- $0.050 / $0.250
novita/fp4fp4 - DeepInfra69.55%Down24h 96.07%
- 7d
- 98.02%
- 30d
- 98.02%
- p50
- 9.41s
- p99
- 12.8s
- Throughput
- 19 tps
- Context
- 131K
- $/Mtok
- $0.037 / $0.170
deepinfra/bf16bf16 - No data24h —
- 7d
- 100%◆ (best)
- 30d
- 100%◆ (best)
- p50
- —
- p99
- —
- Throughput
- —
- Context
- 131K
- $/Mtok
- $0.150 / $0.600
amazon-bedrock/eu-west-1 - Down24h 82.15%
- 7d
- 92.66%
- 30d
- 92.66%
- p50
- 2.68s
- p99
- 17.6s
- Throughput
- 7.0 tps
- Context
- 131K
- $/Mtok
- $0.050 / $0.450
siliconflow/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.