DeepSeek V3.1: 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) | 99.92% | 99.87% | 99.87% | 1.63s | 4.13s | 22 tps | 131K | $0.300 / $0.950 | |
| Healthy | 100%◆ (best) | 100% | 99.99%◆ (best) | 99.99%◆ (best) | 363ms◆ (best) | 951ms◆ (best) | 45 tps | 161K | $0.550 / $1.65 | |
Googlegoogle-vertex/us-west2 | Healthy | 100%◆ (best) | 97.62% | 95.79% | 95.79% | 887ms | 4.73s | 51 tps◆ (best) | 164K | $0.600 / $1.70 |
| Healthy | 100%◆ (best) | 99.99% | 99.99% | 99.99% | 1.71s | 8.66s | 15 tps | 131K | $0.270 / $1.00 | |
| Healthy | 99.92% | 99.85% | 99.54% | 99.54% | 835ms | 8.21s | 10 tps | 164K | $0.250 / $0.950◆ (best) | |
| Degraded | 94.41% | 97.78% | 98.01% | 98.01% | 1.47s | 15.2s | 10 tps | 164K | $0.270 / $1.00 | |
Maramara | Down | 12.86% | 93.88% | 82.96% | 82.96% | 1.05s | 11.5s | 30 tps | 131K | $0.600 / $1.70 |
| Degraded | — | 97.32% | 97.63% | 97.63% | 1.14s | 7.19s | 35 tps | 131K | $0.650 / $1.50 |
- AtlasCloud100%◆ (best)Healthy24h 99.92%
- 7d
- 99.87%
- 30d
- 99.87%
- p50
- 1.63s
- p99
- 4.13s
- Throughput
- 22 tps
- Context
- 131K
- $/Mtok
- $0.300 / $0.950
atlas-cloud/fp8fp8 - CoreWeave100%◆ (best)Healthy24h 100%
- 7d
- 99.99%◆ (best)
- 30d
- 99.99%◆ (best)
- p50
- 363ms◆ (best)
- p99
- 951ms◆ (best)
- Throughput
- 45 tps
- Context
- 161K
- $/Mtok
- $0.550 / $1.65
coreweave/fp8fp8 - Google100%◆ (best)Healthy24h 97.62%
- 7d
- 95.79%
- 30d
- 95.79%
- p50
- 887ms
- p99
- 4.73s
- Throughput
- 51 tps◆ (best)
- Context
- 164K
- $/Mtok
- $0.600 / $1.70
google-vertex/us-west2 - Novita100%◆ (best)Healthy24h 99.99%
- 7d
- 99.99%
- 30d
- 99.99%
- p50
- 1.71s
- p99
- 8.66s
- Throughput
- 15 tps
- Context
- 131K
- $/Mtok
- $0.270 / $1.00
novita/fp8fp8 - DeepInfra99.92%Healthy24h 99.85%
- 7d
- 99.54%
- 30d
- 99.54%
- p50
- 835ms
- p99
- 8.21s
- Throughput
- 10 tps
- Context
- 164K
- $/Mtok
- $0.250 / $0.950◆ (best)
deepinfra/fp4fp4 - SiliconFlow94.41%Degraded24h 97.78%
- 7d
- 98.01%
- 30d
- 98.01%
- p50
- 1.47s
- p99
- 15.2s
- Throughput
- 10 tps
- Context
- 164K
- $/Mtok
- $0.270 / $1.00
siliconflow/fp8fp8 - Mara12.86%Down24h 93.88%
- 7d
- 82.96%
- 30d
- 82.96%
- p50
- 1.05s
- p99
- 11.5s
- Throughput
- 30 tps
- Context
- 131K
- $/Mtok
- $0.600 / $1.70
mara - Degraded24h 97.32%
- 7d
- 97.63%
- 30d
- 97.63%
- p50
- 1.14s
- p99
- 7.19s
- Throughput
- 35 tps
- Context
- 131K
- $/Mtok
- $0.650 / $1.50
sambanova/fp8fp8
Measured 3m 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.