MiniMax M3: 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) | 96.90% | 99.01% | 90.52% | 1.19s | 10.0s | 64 tps | 524K | $0.300 / $1.20 | |
| Healthy | 100%◆ (best) | 99.84% | 99.75%◆ (best) | 98.90% | 1.15s | 7.86s | 35 tps | 524K | $0.280 / $1.10 | |
| Healthy | 100%◆ (best) | 99.39% | 99.39% | 99.51% | 2.43s | 11.5s | 43 tps | 1.0M | $0.240 / $0.960 | |
| Healthy | 100%◆ (best) | 99.65% | 99.51% | 99.63%◆ (best) | 2.01s | 15.0s | 53 tps | 1.0M | $0.300 / $1.20 | |
| Healthy | 100%◆ (best) | 99.81% | 99.58% | 98.44% | 596ms | 2.79s◆ (best) | 175 tps◆ (best) | 1.0M | $0.300 / $1.20 | |
| Healthy | 100%◆ (best) | 99.05% | 98.55% | 99.06% | 1.74s | 10.2s | 47 tps | 1.0M | $0.300 / $1.20 | |
| Healthy | 99.91% | 97.02% | 97.41% | 98.81% | 617ms | 8.77s | 55 tps | 262K | $0.230 / $0.960◆ (best) | |
| Healthy | 99.37% | 99.74% | 99.62% | 99.34% | 1.29s | 13.6s | 65 tps | 524K | $0.300 / $1.20 | |
| Healthy | 98.90% | 94.19% | 96.50% | 95.09% | 1.06s | 9.68s | 83 tps | 524K | $0.300 / $1.20 | |
Togethertogether | Healthy | 97.73% | 99.45% | 99.51% | 99.54% | 1.38s | 10.4s | 22 tps | 524K | $0.300 / $1.20 |
Maramara | No data | — | 95.28% | 95.06% | 92.30% | 1.20s | 8.07s | 133 tps | 1.0M | $0.600 / $2.40 |
| No data | — | 97.86% | 98.33% | 99.00% | 490ms◆ (best) | 4.95s | 146 tps | 1.0M | $0.750 / $3.00 | |
| No data | — | — | — | — | — | — | — | 256K | $0.255 / $1.02 | |
SambaNovasambanova | No data | — | 94.94% | 96.43% | 97.18% | 3.40s | 10.5s | 45 tps | 1.0M | $0.600 / $2.40 |
- AtlasCloud100%◆ (best)Healthy24h 96.90%
- 7d
- 99.01%
- 30d
- 90.52%
- p50
- 1.19s
- p99
- 10.0s
- Throughput
- 64 tps
- Context
- 524K
- $/Mtok
- $0.300 / $1.20
atlas-cloud/fp8fp8 - DeepInfra100%◆ (best)Healthy24h 99.84%
- 7d
- 99.75%◆ (best)
- 30d
- 98.90%
- p50
- 1.15s
- p99
- 7.86s
- Throughput
- 35 tps
- Context
- 524K
- $/Mtok
- $0.280 / $1.10
deepinfra/fp8fp8 - GMICloud100%◆ (best)Healthy24h 99.39%
- 7d
- 99.39%
- 30d
- 99.51%
- p50
- 2.43s
- p99
- 11.5s
- Throughput
- 43 tps
- Context
- 1.0M
- $/Mtok
- $0.240 / $0.960
gmicloud/fp8fp8 - Novita100%◆ (best)Healthy24h 99.65%
- 7d
- 99.51%
- 30d
- 99.63%◆ (best)
- p50
- 2.01s
- p99
- 15.0s
- Throughput
- 53 tps
- Context
- 1.0M
- $/Mtok
- $0.300 / $1.20
novita/fp8fp8 - Parasail100%◆ (best)Healthy24h 99.81%
- 7d
- 99.58%
- 30d
- 98.44%
- p50
- 596ms
- p99
- 2.79s◆ (best)
- Throughput
- 175 tps◆ (best)
- Context
- 1.0M
- $/Mtok
- $0.300 / $1.20
parasail/fp8fp8 - StreamLake100%◆ (best)Healthy24h 99.05%
- 7d
- 98.55%
- 30d
- 99.06%
- p50
- 1.74s
- p99
- 10.2s
- Throughput
- 47 tps
- Context
- 1.0M
- $/Mtok
- $0.300 / $1.20
streamlake/fp8fp8 - CoreWeave99.91%Healthy24h 97.02%
- 7d
- 97.41%
- 30d
- 98.81%
- p50
- 617ms
- p99
- 8.77s
- Throughput
- 55 tps
- Context
- 262K
- $/Mtok
- $0.230 / $0.960◆ (best)
coreweave/fp4fp4 - Minimax99.37%Healthy24h 99.74%
- 7d
- 99.62%
- 30d
- 99.34%
- p50
- 1.29s
- p99
- 13.6s
- Throughput
- 65 tps
- Context
- 524K
- $/Mtok
- $0.300 / $1.20
minimax/fp8fp8 - Venice98.90%Healthy24h 94.19%
- 7d
- 96.50%
- 30d
- 95.09%
- p50
- 1.06s
- p99
- 9.68s
- Throughput
- 83 tps
- Context
- 524K
- $/Mtok
- $0.300 / $1.20
venice/fp8fp8 - Together97.73%Healthy24h 99.45%
- 7d
- 99.51%
- 30d
- 99.54%
- p50
- 1.38s
- p99
- 10.4s
- Throughput
- 22 tps
- Context
- 524K
- $/Mtok
- $0.300 / $1.20
together - Mara—No data24h 95.28%
- 7d
- 95.06%
- 30d
- 92.30%
- p50
- 1.20s
- p99
- 8.07s
- Throughput
- 133 tps
- Context
- 1.0M
- $/Mtok
- $0.600 / $2.40
mara - No data24h 97.86%
- 7d
- 98.33%
- 30d
- 99.00%
- p50
- 490ms◆ (best)
- p99
- 4.95s
- Throughput
- 146 tps
- Context
- 1.0M
- $/Mtok
- $0.750 / $3.00
modelrun/fp4fp4 - No data24h —
- 7d
- —
- 30d
- —
- p50
- —
- p99
- —
- Throughput
- —
- Context
- 256K
- $/Mtok
- $0.255 / $1.02
morph/fp4fp4 - No data24h 94.94%
- 7d
- 96.43%
- 30d
- 97.18%
- p50
- 3.40s
- p99
- 10.5s
- Throughput
- 45 tps
- Context
- 1.0M
- $/Mtok
- $0.600 / $2.40
sambanova
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.