Kimi K3: 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 | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
Fireworksfireworks/fast | Healthy | 100%◆ (best) | 95.10% | 98.58% | 99.03% | 700ms | 4.24s | 97 tps◆ (best) | 1.0M | $4.50 / $22.50 |
Fireworksfireworks/us | Healthy | 100%◆ (best) | 99.35% | 99.45% | 99.21% | 2.99s | 12.7s | 50 tps | 1.0M | $4.50 / $22.50 |
Makoramakora | Healthy | 100%◆ (best) | 95.85% | 97.78% | 97.06% | 1.18s | 24.3s | 53 tps | 1.0M | $1.53 / $12.75 |
| Healthy | 100%◆ (best) | 98.61% | 98.87% | 99.56% | 1.89s | 26.2s | 66 tps | 1.0M | $3.00 / $15.00 | |
| Healthy | 100%◆ (best) | 99.99% | 99.55% | 99.67% | 4.58s | 25.4s | 31 tps | 1.0M | $3.00 / $15.00 | |
| Healthy | 100%◆ (best) | 99.09% | 98.28% | 98.97% | 1.29s | 5.62s | 39 tps | 1.0M | $0.375 / $12.83 | |
| Healthy | 100%◆ (best) | 99.69% | 90.45% | 96.74% | 1.26s | 14.9s | 63 tps | 1.0M | $0.365 / $13.00◆ (best) | |
| Healthy | 100%◆ (best) | 99.81% | 99.77% | 99.64% | 1.53s | 6.87s | 75 tps | 1.0M | $2.80 / $14.00 | |
Waferwafer | Healthy | 100%◆ (best) | 99.41% | 99.79%◆ (best) | 99.55% | 956ms | 18.4s | 27 tps | 1.0M | $0.790 / $14.00 |
Waferwafer/us | Healthy | 100%◆ (best) | 98.69% | 99.55% | 99.55% | 739ms | 28.0s | 36 tps | 1.0M | $2.80 / $14.00 |
Togethertogether | Healthy | 99.90% | 99.26% | 99.40% | 99.34% | 1.07s | 9.85s | 49 tps | 1.0M | $2.70 / $13.50 |
| Healthy | 99.71% | 99.90% | 99.68% | 99.13% | 895ms | 6.39s | 64 tps | 1.0M | $0.790 / $14.00 | |
Fireworksfireworks | Healthy | 99.68% | 99.93% | 99.61% | 99.80%◆ (best) | 1.53s | 6.97s | 28 tps | 1.0M | $3.00 / $15.00 |
| Healthy | 99.57% | 98.34% | 98.59% | 97.40% | 1.07s | 11.6s | 80 tps | 1.0M | $3.00 / $15.00 | |
| Healthy | 98.48% | 99.76% | 99.35% | 99.35% | 896ms | 4.14s◆ (best) | 75 tps | 250K | $3.50 / $15.00 | |
Phalaphala | Degraded | 91.37% | 96.50% | 98.08% | 98.84% | 4.24s | 49.1s | 22 tps | 1.0M | $1.95 / $9.75 |
Amazon Bedrockamazon-bedrock/us-east-2 | Degraded | 90.00% | 94.61% | 96.02% | 96.02% | 1.96s | 33.4s | 75 tps | 1.0M | $3.00 / $15.00 |
| No data | — | 98.52% | 99.73% | 99.73% | 656ms◆ (best) | 5.64s | 71 tps | 1.0M | $1.30 / $14.00 | |
Alibabaalibaba | No data | — | 98.98% | 99.04% | 98.94% | 1.42s | 7.95s | 50 tps | 1.0M | $3.45 / $17.25 |
| Degraded | — | 97.57% | 98.39% | 98.59% | 2.63s | 21.7s | 60 tps | 1.0M | $3.00 / $15.00 | |
| No data | — | 97.29% | 94.39% | 97.42% | 2.55s | 10.2s | 27 tps | 1.0M | $3.00 / $15.00 | |
| No data | — | 70.58% | 90.36% | 90.32% | 1.66s | 10.2s | 40 tps | 1.0M | $2.52 / $12.60 | |
| No data | — | — | 98.78% | 98.06% | — | — | — | 1.0M | $2.85 / $14.25 | |
| No data | — | 99.36% | 99.46% | 99.46% | 891ms | 10.5s | 69 tps | 1.0M | $2.85 / $14.25 | |
DigitalOceandigitalocean | No data | — | 99.94% | 99.51% | 99.41% | 1.09s | 9.67s | 38 tps | 1.0M | $2.55 / $12.95 |
InferenceNetinference-net | No data | — | — | — | 96.45% | — | — | — | 1.0M | $2.10 / $10.95 |
Morphmorph | No data | — | — | — | 99.41% | — | — | — | 1.0M | $2.10 / $11.74 |
Morphmorph/fast | No data | — | — | — | 97.50% | — | — | — | 1.0M | $6.00 / $22.50 |
| No data | — | — | — | 98.41% | — | — | — | 1.0M | $2.50 / $14.00 |
- Fireworks100%◆ (best)Healthy24h 95.10%
- 7d
- 98.58%
- 30d
- 99.03%
- p50
- 700ms
- p99
- 4.24s
- Throughput
- 97 tps◆ (best)
- Context
- 1.0M
- $/Mtok
- $4.50 / $22.50
fireworks/fast - Fireworks100%◆ (best)Healthy24h 99.35%
- 7d
- 99.45%
- 30d
- 99.21%
- p50
- 2.99s
- p99
- 12.7s
- Throughput
- 50 tps
- Context
- 1.0M
- $/Mtok
- $4.50 / $22.50
fireworks/us - Makora100%◆ (best)Healthy24h 95.85%
- 7d
- 97.78%
- 30d
- 97.06%
- p50
- 1.18s
- p99
- 24.3s
- Throughput
- 53 tps
- Context
- 1.0M
- $/Mtok
- $1.53 / $12.75
makora - Modal100%◆ (best)Healthy24h 98.61%
- 7d
- 98.87%
- 30d
- 99.56%
- p50
- 1.89s
- p99
- 26.2s
- Throughput
- 66 tps
- Context
- 1.0M
- $/Mtok
- $3.00 / $15.00
modal/mxfp4mxfp4 - Moonshot AI100%◆ (best)Healthy24h 99.99%
- 7d
- 99.55%
- 30d
- 99.67%
- p50
- 4.58s
- p99
- 25.4s
- Throughput
- 31 tps
- Context
- 1.0M
- $/Mtok
- $3.00 / $15.00
moonshotai/mxfp4mxfp4 - Morph100%◆ (best)Healthy24h 99.09%
- 7d
- 98.28%
- 30d
- 98.97%
- p50
- 1.29s
- p99
- 5.62s
- Throughput
- 39 tps
- Context
- 1.0M
- $/Mtok
- $0.375 / $12.83
morph/fp8fp8 - Relace100%◆ (best)Healthy24h 99.69%
- 7d
- 90.45%
- 30d
- 96.74%
- p50
- 1.26s
- p99
- 14.9s
- Throughput
- 63 tps
- Context
- 1.0M
- $/Mtok
- $0.365 / $13.00◆ (best)
relace/fp4fp4 - Sail Research100%◆ (best)Healthy24h 99.81%
- 7d
- 99.77%
- 30d
- 99.64%
- p50
- 1.53s
- p99
- 6.87s
- Throughput
- 75 tps
- Context
- 1.0M
- $/Mtok
- $2.80 / $14.00
sail-research/fp4fp4 - Wafer100%◆ (best)Healthy24h 99.41%
- 7d
- 99.79%◆ (best)
- 30d
- 99.55%
- p50
- 956ms
- p99
- 18.4s
- Throughput
- 27 tps
- Context
- 1.0M
- $/Mtok
- $0.790 / $14.00
wafer - Wafer100%◆ (best)Healthy24h 98.69%
- 7d
- 99.55%
- 30d
- 99.55%
- p50
- 739ms
- p99
- 28.0s
- Throughput
- 36 tps
- Context
- 1.0M
- $/Mtok
- $2.80 / $14.00
wafer/us - Together99.90%Healthy24h 99.26%
- 7d
- 99.40%
- 30d
- 99.34%
- p50
- 1.07s
- p99
- 9.85s
- Throughput
- 49 tps
- Context
- 1.0M
- $/Mtok
- $2.70 / $13.50
together - InferenceNet99.71%Healthy24h 99.90%
- 7d
- 99.68%
- 30d
- 99.13%
- p50
- 895ms
- p99
- 6.39s
- Throughput
- 64 tps
- Context
- 1.0M
- $/Mtok
- $0.790 / $14.00
inference-net/fp4fp4 - Fireworks99.68%Healthy24h 99.93%
- 7d
- 99.61%
- 30d
- 99.80%◆ (best)
- p50
- 1.53s
- p99
- 6.97s
- Throughput
- 28 tps
- Context
- 1.0M
- $/Mtok
- $3.00 / $15.00
fireworks - Parasail99.57%Healthy24h 98.34%
- 7d
- 98.59%
- 30d
- 97.40%
- p50
- 1.07s
- p99
- 11.6s
- Throughput
- 80 tps
- Context
- 1.0M
- $/Mtok
- $3.00 / $15.00
parasail/fp4fp4 - InferenceNet98.48%Healthy24h 99.76%
- 7d
- 99.35%
- 30d
- 99.35%
- p50
- 896ms
- p99
- 4.14s◆ (best)
- Throughput
- 75 tps
- Context
- 250K
- $/Mtok
- $3.50 / $15.00
inference-net/fastfp4 - Phala91.37%Degraded24h 96.50%
- 7d
- 98.08%
- 30d
- 98.84%
- p50
- 4.24s
- p99
- 49.1s
- Throughput
- 22 tps
- Context
- 1.0M
- $/Mtok
- $1.95 / $9.75
phala - Amazon Bedrock90.00%Degraded24h 94.61%
- 7d
- 96.02%
- 30d
- 96.02%
- p50
- 1.96s
- p99
- 33.4s
- Throughput
- 75 tps
- Context
- 1.0M
- $/Mtok
- $3.00 / $15.00
amazon-bedrock/us-east-2 - No data24h 98.52%
- 7d
- 99.73%
- 30d
- 99.73%
- p50
- 656ms◆ (best)
- p99
- 5.64s
- Throughput
- 71 tps
- Context
- 1.0M
- $/Mtok
- $1.30 / $14.00
akashml/fp4fp4 - No data24h 98.98%
- 7d
- 99.04%
- 30d
- 98.94%
- p50
- 1.42s
- p99
- 7.95s
- Throughput
- 50 tps
- Context
- 1.0M
- $/Mtok
- $3.45 / $17.25
alibaba - Degraded24h 97.57%
- 7d
- 98.39%
- 30d
- 98.59%
- p50
- 2.63s
- p99
- 21.7s
- Throughput
- 60 tps
- Context
- 1.0M
- $/Mtok
- $3.00 / $15.00
baseten/fp8fp8 - No data24h 97.29%
- 7d
- 94.39%
- 30d
- 97.42%
- p50
- 2.55s
- p99
- 10.2s
- Throughput
- 27 tps
- Context
- 1.0M
- $/Mtok
- $3.00 / $15.00
chutes/mxfp4mxfp4 - No data24h 70.58%
- 7d
- 90.36%
- 30d
- 90.32%
- p50
- 1.66s
- p99
- 10.2s
- Throughput
- 40 tps
- Context
- 1.0M
- $/Mtok
- $2.52 / $12.60
decart/mxfp4mxfp4 - No data24h —
- 7d
- 98.78%
- 30d
- 98.06%
- p50
- —
- p99
- —
- Throughput
- —
- Context
- 1.0M
- $/Mtok
- $2.85 / $14.25
deepinfra/bf16bf16 - No data24h 99.36%
- 7d
- 99.46%
- 30d
- 99.46%
- p50
- 891ms
- p99
- 10.5s
- Throughput
- 69 tps
- Context
- 1.0M
- $/Mtok
- $2.85 / $14.25
deepinfra/mxfp4mxfp4 - No data24h 99.94%
- 7d
- 99.51%
- 30d
- 99.41%
- p50
- 1.09s
- p99
- 9.67s
- Throughput
- 38 tps
- Context
- 1.0M
- $/Mtok
- $2.55 / $12.95
digitalocean - No data24h —
- 7d
- —
- 30d
- 96.45%
- p50
- —
- p99
- —
- Throughput
- —
- Context
- 1.0M
- $/Mtok
- $2.10 / $10.95
inference-net - No data24h —
- 7d
- —
- 30d
- 99.41%
- p50
- —
- p99
- —
- Throughput
- —
- Context
- 1.0M
- $/Mtok
- $2.10 / $11.74
morph - No data24h —
- 7d
- —
- 30d
- 97.50%
- p50
- —
- p99
- —
- Throughput
- —
- Context
- 1.0M
- $/Mtok
- $6.00 / $22.50
morph/fast - No data24h —
- 7d
- —
- 30d
- 98.41%
- p50
- —
- p99
- —
- Throughput
- —
- Context
- 1.0M
- $/Mtok
- $2.50 / $14.00
morph/fp4fp4
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, averaged over the hour.
Throughput
Median tokens per second.