Kimi K3
moonshotai/kimi-k3
Compare with Kimi K2.7 Code, Kimi K2.6, Kimi K2.5
CompareKimi K3
| State | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
Fireworksfireworks | Healthy | 100% | 98.41% | 98.92% | 98.92% | 3.04s | 25.4s | 36 tps | 1.0M | $3.00 / $15.00 |
| Healthy | 99.77% | 99.55% | 99.65% | 99.65% | 1.83s | 17.7s | 78 tps | 1.0M | $3.00 / $15.00 | |
| Healthy | 99.37% | 99.28% | 97.14% | 97.14% | 1.55s | 15.5s | 21 tps | 1.0M | $3.00 / $15.00 | |
DigitalOceandigitalocean | Healthy | 99.27% | 99.45% | 99.30% | 99.30% | 2.97s | 31.2s | 23 tps | 1.0M | $2.85 / $14.25 |
| Healthy | 98.97% | 94.89% | 87.79% | 87.79% | 3.27s | 46.9s | 23 tps | 1.0M | $2.80 / $14.00 | |
| Healthy | 97.65% | 97.48% | 98.50% | 98.50% | 4.15s | 87.2s | 26 tps | 975K | $2.60 / $13.00 | |
| Degraded | 97.59% | 98.67% | 99.30% | 99.30% | 6.25s | 48.4s | 21 tps | 1.0M | $3.00 / $15.00 | |
Togethertogether | Degraded | 93.25% | 98.95% | 98.77% | 98.77% | 2.10s | 19.6s | 46 tps | 1.0M | $3.00 / $15.00 |
Fireworksfireworks/us | Degraded | 91.67% | 92.81% | 95.41% | 95.41% | 827ms | 10.0s | 64 tps | 1.0M | $3.30 / $16.50 |
| No data | — | 98.40% | 98.19% | 98.19% | 1.13s | 19.7s | 71 tps | 1.0M | $3.00 / $15.00 | |
| Down | — | 92.22% | 89.41% | 89.41% | 934ms | 26.1s | 18 tps | 1.0M | $2.85 / $14.25 | |
| No data | — | 84.16% | 82.82% | 82.82% | 5.22s | 12.3s | 24 tps | 1.0M | $6.00 / $22.50 | |
Phalaphala | Down | — | 96.26% | 95.10% | 95.10% | 3.11s | 25.9s | 38 tps | 1.0M | $3.00 / $15.00 |
Waferwafer | No data | — | 97.35% | 96.96% | 96.96% | — | — | — | 1.0M | $3.00 / $15.00 |
- Fireworks100%Healthy24h 98.41%
- 7d
- 98.92%
- 30d
- 98.92%
- p50
- 3.04s
- p99
- 25.4s
- Throughput
- 36 tps
- Context
- 1.0M
- $/Mtok
- $3.00 / $15.00
fireworks - Modal99.77%Healthy24h 99.55%
- 7d
- 99.65%
- 30d
- 99.65%
- p50
- 1.83s
- p99
- 17.7s
- Throughput
- 78 tps
- Context
- 1.0M
- $/Mtok
- $3.00 / $15.00
modal/mxfp4mxfp4 - Chutes99.37%Healthy24h 99.28%
- 7d
- 97.14%
- 30d
- 97.14%
- p50
- 1.55s
- p99
- 15.5s
- Throughput
- 21 tps
- Context
- 1.0M
- $/Mtok
- $3.00 / $15.00
chutes/mxfp4mxfp4 - DigitalOcean99.27%Healthy24h 99.45%
- 7d
- 99.30%
- 30d
- 99.30%
- p50
- 2.97s
- p99
- 31.2s
- Throughput
- 23 tps
- Context
- 1.0M
- $/Mtok
- $2.85 / $14.25
digitalocean - Morph98.97%Healthy24h 94.89%
- 7d
- 87.79%
- 30d
- 87.79%
- p50
- 3.27s
- p99
- 46.9s
- Throughput
- 23 tps
- Context
- 1.0M
- $/Mtok
- $2.80 / $14.00
morph/fp4fp4 - Sail Research97.65%Healthy24h 97.48%
- 7d
- 98.50%
- 30d
- 98.50%
- p50
- 4.15s
- p99
- 87.2s
- Throughput
- 26 tps
- Context
- 975K
- $/Mtok
- $2.60 / $13.00
sail-research/fp4fp4 - Moonshot AI97.59%Degraded24h 98.67%
- 7d
- 99.30%
- 30d
- 99.30%
- p50
- 6.25s
- p99
- 48.4s
- Throughput
- 21 tps
- Context
- 1.0M
- $/Mtok
- $3.00 / $15.00
moonshotai/mxfp4mxfp4 - Together93.25%Degraded24h 98.95%
- 7d
- 98.77%
- 30d
- 98.77%
- p50
- 2.10s
- p99
- 19.6s
- Throughput
- 46 tps
- Context
- 1.0M
- $/Mtok
- $3.00 / $15.00
together - Fireworks91.67%Degraded24h 92.81%
- 7d
- 95.41%
- 30d
- 95.41%
- p50
- 827ms
- p99
- 10.0s
- Throughput
- 64 tps
- Context
- 1.0M
- $/Mtok
- $3.30 / $16.50
fireworks/us - No data24h 98.40%
- 7d
- 98.19%
- 30d
- 98.19%
- p50
- 1.13s
- p99
- 19.7s
- Throughput
- 71 tps
- Context
- 1.0M
- $/Mtok
- $3.00 / $15.00
baseten/fp8fp8 - Down24h 92.22%
- 7d
- 89.41%
- 30d
- 89.41%
- p50
- 934ms
- p99
- 26.1s
- Throughput
- 18 tps
- Context
- 1.0M
- $/Mtok
- $2.85 / $14.25
deepinfra/bf16bf16 - No data24h 84.16%
- 7d
- 82.82%
- 30d
- 82.82%
- p50
- 5.22s
- p99
- 12.3s
- Throughput
- 24 tps
- Context
- 1.0M
- $/Mtok
- $6.00 / $22.50
morph/fastfp4 - Down24h 96.26%
- 7d
- 95.10%
- 30d
- 95.10%
- p50
- 3.11s
- p99
- 25.9s
- Throughput
- 38 tps
- Context
- 1.0M
- $/Mtok
- $3.00 / $15.00
phala - No data24h 97.35%
- 7d
- 96.96%
- 30d
- 96.96%
- p50
- —
- p99
- —
- Throughput
- —
- Context
- 1.0M
- $/Mtok
- $3.00 / $15.00
wafer
Measured 1m ago. Latency and throughput are OpenRouter's rolling 30-minute windows.
History
3 endpoints reporting no data are omitted from the charts.
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.