DeepSeek V4 Pro 0813: 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) | 99.04% | 98.67% | 98.67% | 1.18s | 26.0s | 94 tps | 1.0M | $1.32 / $3.96 | |
DeepSeekdeepseek | Healthy | 100%◆ (best) | 99.95% | 86.81% | 86.81% | 1.39s | 9.34s | 33 tps | 1.0M | $0.660 / $1.98 |
Fireworksfireworks | Healthy | 100%◆ (best) | 98.47% | 99.23% | 99.23% | 1.29s | 7.46s | 45 tps | 1.0M | $1.32 / $3.96 |
| Healthy | 100%◆ (best) | 99.59% | 99.67%◆ (best) | 99.67%◆ (best) | 3.47s | 25.1s | 44 tps | 1.0M | $1.19 / $3.56 | |
| Healthy | 100%◆ (best) | 99.47% | 99.03% | 99.03% | 1.33s | 8.89s | 52 tps | 1.0M | $1.32 / $3.96 | |
| Healthy | 100%◆ (best) | 99.74% | 99.50% | 99.50% | 868ms | 4.44s◆ (best) | 108 tps◆ (best) | 1.0M | $1.32 / $3.96 | |
StreamLakestreamlake | Healthy | 100%◆ (best) | 98.24% | 98.72% | 98.72% | 3.12s | 103s | 33 tps | 1.0M | $1.32 / $3.96 |
Togethertogether | Healthy | 100%◆ (best) | 97.47% | 97.17% | 97.17% | 1.14s | 12.0s | 82 tps | 1.0M | $1.32 / $3.96 |
Alibabaalibaba | Healthy | 99.91% | 99.83% | 99.62% | 99.62% | 1.58s | 17.3s | 52 tps | 1.0M | $0.603 / $1.81◆ (best) |
| Healthy | 98.84% | 96.71% | 95.97% | 95.97% | 699ms◆ (best) | 8.17s | 71 tps | 1.0M | $1.32 / $3.96 | |
Cloudflarecloudflare | Degraded | 96.97% | 97.87% | 97.21% | 97.21% | 4.85s | 86.0s | 39 tps | 1.0M | $1.32 / $3.96 |
DigitalOceandigitalocean | Degraded | — | 97.80% | 97.04% | 97.04% | 1.74s | 10.8s | 36 tps | 1.0M | $1.32 / $3.96 |
| No data | — | 98.33% | 99.12% | 99.12% | 2.17s | 60.4s | 40 tps | 1.0M | $1.32 / $3.96 |
- BaseTen100%◆ (best)Healthy24h 99.04%
- 7d
- 98.67%
- 30d
- 98.67%
- p50
- 1.18s
- p99
- 26.0s
- Throughput
- 94 tps
- Context
- 1.0M
- $/Mtok
- $1.32 / $3.96
baseten/fp4fp4 - DeepSeek100%◆ (best)Healthy24h 99.95%
- 7d
- 86.81%
- 30d
- 86.81%
- p50
- 1.39s
- p99
- 9.34s
- Throughput
- 33 tps
- Context
- 1.0M
- $/Mtok
- $0.660 / $1.98
deepseek - Fireworks100%◆ (best)Healthy24h 98.47%
- 7d
- 99.23%
- 30d
- 99.23%
- p50
- 1.29s
- p99
- 7.46s
- Throughput
- 45 tps
- Context
- 1.0M
- $/Mtok
- $1.32 / $3.96
fireworks - GMICloud100%◆ (best)Healthy24h 99.59%
- 7d
- 99.67%◆ (best)
- 30d
- 99.67%◆ (best)
- p50
- 3.47s
- p99
- 25.1s
- Throughput
- 44 tps
- Context
- 1.0M
- $/Mtok
- $1.19 / $3.56
gmicloud/fp8fp8 - Novita100%◆ (best)Healthy24h 99.47%
- 7d
- 99.03%
- 30d
- 99.03%
- p50
- 1.33s
- p99
- 8.89s
- Throughput
- 52 tps
- Context
- 1.0M
- $/Mtok
- $1.32 / $3.96
novita/fp8fp8 - Sail Research100%◆ (best)Healthy24h 99.74%
- 7d
- 99.50%
- 30d
- 99.50%
- p50
- 868ms
- p99
- 4.44s◆ (best)
- Throughput
- 108 tps◆ (best)
- Context
- 1.0M
- $/Mtok
- $1.32 / $3.96
sail-research/fp4fp4 - StreamLake100%◆ (best)Healthy24h 98.24%
- 7d
- 98.72%
- 30d
- 98.72%
- p50
- 3.12s
- p99
- 103s
- Throughput
- 33 tps
- Context
- 1.0M
- $/Mtok
- $1.32 / $3.96
streamlake - Together100%◆ (best)Healthy24h 97.47%
- 7d
- 97.17%
- 30d
- 97.17%
- p50
- 1.14s
- p99
- 12.0s
- Throughput
- 82 tps
- Context
- 1.0M
- $/Mtok
- $1.32 / $3.96
together - Alibaba99.91%Healthy24h 99.83%
- 7d
- 99.62%
- 30d
- 99.62%
- p50
- 1.58s
- p99
- 17.3s
- Throughput
- 52 tps
- Context
- 1.0M
- $/Mtok
- $0.603 / $1.81◆ (best)
alibaba - Parasail98.84%Healthy24h 96.71%
- 7d
- 95.97%
- 30d
- 95.97%
- p50
- 699ms◆ (best)
- p99
- 8.17s
- Throughput
- 71 tps
- Context
- 1.0M
- $/Mtok
- $1.32 / $3.96
parasail/fp8fp8 - Cloudflare96.97%Degraded24h 97.87%
- 7d
- 97.21%
- 30d
- 97.21%
- p50
- 4.85s
- p99
- 86.0s
- Throughput
- 39 tps
- Context
- 1.0M
- $/Mtok
- $1.32 / $3.96
cloudflare - Degraded24h 97.80%
- 7d
- 97.04%
- 30d
- 97.04%
- p50
- 1.74s
- p99
- 10.8s
- Throughput
- 36 tps
- Context
- 1.0M
- $/Mtok
- $1.32 / $3.96
digitalocean - No data24h 98.33%
- 7d
- 99.12%
- 30d
- 99.12%
- p50
- 2.17s
- p99
- 60.4s
- Throughput
- 40 tps
- Context
- 1.0M
- $/Mtok
- $1.32 / $3.96
siliconflow/fp8fp8
Measured 46s 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.