Gemma 4 31B: 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
Cerebrascerebras/fp16fp16	Healthy	100%◆ (best)	100%	100%◆ (best)	100%◆ (best)	429ms	6.88s	131 tps◆ (best)	131K	$0.990 / $1.49
DeepInfradeepinfra/fp8fp8	Healthy	100%◆ (best)	98.41%	90.99%	90.99%	444ms	5.53s	42 tps	262K	$0.130 / $0.380
ModelRunmodelrun/fp4fp4	Healthy	100%◆ (best)	99.92%	99.66%	99.66%	225ms◆ (best)	3.54s◆ (best)	74 tps	262K	$0.750 / $1.00
SambaNovasambanova	Degraded	100%◆ (best)	97.99%	97.09%	97.09%	3.67s	31.3s	18 tps	262K	$0.380 / $1.15
Togethertogether	Healthy	100%◆ (best)	92.36%	95.49%	95.49%	702ms	15.2s	31 tps	262K	$0.280 / $0.860
Friendlifriendli	Healthy	99.87%	99.83%	99.59%	99.59%	2.29s	12.1s	39 tps	262K	$0.140 / $0.400
Venicevenice/bf16bf16	Healthy	99.78%	99.76%	99.74%	99.74%	1.11s	18.7s	36 tps	256K	$0.120 / $0.360
DeepInfradeepinfra/turbofp4	Healthy	99.76%	99.84%	99.54%	99.54%	776ms	5.89s	16 tps	262K	$0.090 / $0.340
OpenInferenceopen-inference/bf16bf16	Healthy	99.75%	97.72%	97.58%	97.58%	1.02s	36.6s	18 tps	262K	$0.080 / $0.350◆ (best)
Novitanovita/bf16bf16	Healthy	99.62%	99.02%	80.00%	80.00%	977ms	8.68s	13 tps	262K	$0.140 / $0.400
Parasailparasail/fp8fp8	Healthy	99.56%	99.59%	99.29%	99.29%	714ms	9.70s	20 tps	262K	$0.150 / $0.400
CoreWeavecoreweave/bf16bf16	Degraded	98.49%	98.20%	98.51%	98.51%	1.36s	30.8s	27 tps	262K	$0.100 / $0.340
Crusoecrusoe	Degraded	93.53%	98.81%	98.66%	98.66%	1.18s	27.3s	21 tps	262K	$0.140 / $0.400
Chuteschutes/fp4fp4	Degraded	90.09%	89.19%	64.54%	64.54%	2.21s	23.2s	12 tps	131K	$0.120 / $0.370
DeepInfradeepinfra/ultrafp8	Down	87.14%	79.17%	88.70%	88.70%	5.12s	20.2s	16 tps	131K	$0.270 / $0.760
SiliconFlowsiliconflow/fp8fp8	Down	75.12%	84.55%	94.20%	94.20%	2.36s	41.8s	23 tps	262K	$0.130 / $0.400
Phalaphala	Down	7.39%	75.22%	85.43%	85.43%	1.38s	14.3s	43 tps	262K	$0.150 / $0.460

Cerebras100%◆ (best)
Healthy24h 100%
7d
100%◆ (best)
30d
100%◆ (best)
p50
429ms
p99
6.88s
Throughput
131 tps◆ (best)
Context
131K
$/Mtok
$0.990 / $1.49
cerebras/fp16fp16
DeepInfra100%◆ (best)
Healthy24h 98.41%
7d
90.99%
30d
90.99%
p50
444ms
p99
5.53s
Throughput
42 tps
Context
262K
$/Mtok
$0.130 / $0.380
deepinfra/fp8fp8
ModelRun100%◆ (best)
Healthy24h 99.92%
7d
99.66%
30d
99.66%
p50
225ms◆ (best)
p99
3.54s◆ (best)
Throughput
74 tps
Context
262K
$/Mtok
$0.750 / $1.00
modelrun/fp4fp4
SambaNova100%◆ (best)
Degraded24h 97.99%
7d
97.09%
30d
97.09%
p50
3.67s
p99
31.3s
Throughput
18 tps
Context
262K
$/Mtok
$0.380 / $1.15
sambanova
Together100%◆ (best)
Healthy24h 92.36%
7d
95.49%
30d
95.49%
p50
702ms
p99
15.2s
Throughput
31 tps
Context
262K
$/Mtok
$0.280 / $0.860
together
Friendli99.87%
Healthy24h 99.83%
7d
99.59%
30d
99.59%
p50
2.29s
p99
12.1s
Throughput
39 tps
Context
262K
$/Mtok
$0.140 / $0.400
friendli
Venice99.78%
Healthy24h 99.76%
7d
99.74%
30d
99.74%
p50
1.11s
p99
18.7s
Throughput
36 tps
Context
256K
$/Mtok
$0.120 / $0.360
venice/bf16bf16
DeepInfra99.76%
Healthy24h 99.84%
7d
99.54%
30d
99.54%
p50
776ms
p99
5.89s
Throughput
16 tps
Context
262K
$/Mtok
$0.090 / $0.340
deepinfra/turbofp4
OpenInference99.75%
Healthy24h 97.72%
7d
97.58%
30d
97.58%
p50
1.02s
p99
36.6s
Throughput
18 tps
Context
262K
$/Mtok
$0.080 / $0.350◆ (best)
open-inference/bf16bf16
Novita99.62%
Healthy24h 99.02%
7d
80.00%
30d
80.00%
p50
977ms
p99
8.68s
Throughput
13 tps
Context
262K
$/Mtok
$0.140 / $0.400
novita/bf16bf16
Parasail99.56%
Healthy24h 99.59%
7d
99.29%
30d
99.29%
p50
714ms
p99
9.70s
Throughput
20 tps
Context
262K
$/Mtok
$0.150 / $0.400
parasail/fp8fp8
CoreWeave98.49%
Degraded24h 98.20%
7d
98.51%
30d
98.51%
p50
1.36s
p99
30.8s
Throughput
27 tps
Context
262K
$/Mtok
$0.100 / $0.340
coreweave/bf16bf16
Crusoe93.53%
Degraded24h 98.81%
7d
98.66%
30d
98.66%
p50
1.18s
p99
27.3s
Throughput
21 tps
Context
262K
$/Mtok
$0.140 / $0.400
crusoe
Chutes90.09%
Degraded24h 89.19%
7d
64.54%
30d
64.54%
p50
2.21s
p99
23.2s
Throughput
12 tps
Context
131K
$/Mtok
$0.120 / $0.370
chutes/fp4fp4
DeepInfra87.14%
Down24h 79.17%
7d
88.70%
30d
88.70%
p50
5.12s
p99
20.2s
Throughput
16 tps
Context
131K
$/Mtok
$0.270 / $0.760
deepinfra/ultrafp8
SiliconFlow75.12%
Down24h 84.55%
7d
94.20%
30d
94.20%
p50
2.36s
p99
41.8s
Throughput
23 tps
Context
262K
$/Mtok
$0.130 / $0.400
siliconflow/fp8fp8
Phala7.39%
Down24h 75.22%
7d
85.43%
30d
85.43%
p50
1.38s
p99
14.3s
Throughput
43 tps
Context
262K
$/Mtok
$0.150 / $0.460
phala

Measured 2m 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.

Throughput

Median tokens per second.