Llama 3.3 70B Instruct: 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
CoreWeavecoreweave/fp16fp16	Healthy	100%◆ (best)	98.64%	97.83%	97.83%	167ms◆ (best)	1.46s	50 tps	128K	$0.710 / $0.710
Crusoecrusoe/bf16bf16	Healthy	100%◆ (best)	99.74%	99.08%	99.08%	380ms	2.51s	47 tps	131K	$0.250 / $0.750
Googlegoogle-vertex/us-central1	Healthy	100%◆ (best)	99.99%	100%◆ (best)	100%◆ (best)	389ms	2.09s	39 tps	128K	$0.720 / $0.720
Groqgroq	Healthy	100%◆ (best)	99.91%	99.57%	99.57%	198ms	616ms◆ (best)	18 tps	131K	$0.590 / $0.790
AkashMLakashml/fp8fp8	Healthy	99.64%	97.01%	97.28%	97.28%	777ms	5.11s	22 tps	131K	$0.200 / $0.520
Parasailparasail/fp8fp8	Healthy	99.33%	98.31%	97.90%	97.90%	625ms	5.92s	35 tps	131K	$0.220 / $0.500
Cloudflarecloudflare/fp8fp8	Healthy	98.51%	95.94%	94.55%	94.55%	682ms	5.58s	29 tps	24K	$0.293 / $2.25
Novitanovita/bf16bf16	Healthy	97.99%	94.77%	94.32%	94.32%	792ms	6.61s	28 tps	12K	$0.135 / $0.400
Googlegoogle-vertex	Healthy	97.78%	98.81%	98.30%	98.30%	393ms	2.01s	38 tps	128K	$0.720 / $0.720
SambaNovasambanova-turbo	Degraded	96.71%	96.75%	95.41%	95.41%	672ms	5.68s	79 tps◆ (best)	131K	$0.450 / $0.900
DeepInfradeepinfra/turbofp8	Degraded	96.02%	96.82%	96.49%	96.49%	841ms	7.50s	7.0 tps	131K	$0.100 / $0.320◆ (best)
Nebiusnebius/fp8fp8	Down	—	41.46%	42.69%	42.69%	1.86s	14.3s	7.0 tps	131K	$0.130 / $0.400
Togethertogether	No data	—	98.50%	97.76%	97.76%	680ms	3.14s	30 tps	131K	$1.04 / $1.04

CoreWeave100%◆ (best)
Healthy24h 98.64%
7d
97.83%
30d
97.83%
p50
167ms◆ (best)
p99
1.46s
Throughput
50 tps
Context
128K
$/Mtok
$0.710 / $0.710
coreweave/fp16fp16
Crusoe100%◆ (best)
Healthy24h 99.74%
7d
99.08%
30d
99.08%
p50
380ms
p99
2.51s
Throughput
47 tps
Context
131K
$/Mtok
$0.250 / $0.750
crusoe/bf16bf16
Google100%◆ (best)
Healthy24h 99.99%
7d
100%◆ (best)
30d
100%◆ (best)
p50
389ms
p99
2.09s
Throughput
39 tps
Context
128K
$/Mtok
$0.720 / $0.720
google-vertex/us-central1
Groq100%◆ (best)
Healthy24h 99.91%
7d
99.57%
30d
99.57%
p50
198ms
p99
616ms◆ (best)
Throughput
18 tps
Context
131K
$/Mtok
$0.590 / $0.790
groq
AkashML99.64%
Healthy24h 97.01%
7d
97.28%
30d
97.28%
p50
777ms
p99
5.11s
Throughput
22 tps
Context
131K
$/Mtok
$0.200 / $0.520
akashml/fp8fp8
Parasail99.33%
Healthy24h 98.31%
7d
97.90%
30d
97.90%
p50
625ms
p99
5.92s
Throughput
35 tps
Context
131K
$/Mtok
$0.220 / $0.500
parasail/fp8fp8
Cloudflare98.51%
Healthy24h 95.94%
7d
94.55%
30d
94.55%
p50
682ms
p99
5.58s
Throughput
29 tps
Context
24K
$/Mtok
$0.293 / $2.25
cloudflare/fp8fp8
Novita97.99%
Healthy24h 94.77%
7d
94.32%
30d
94.32%
p50
792ms
p99
6.61s
Throughput
28 tps
Context
12K
$/Mtok
$0.135 / $0.400
novita/bf16bf16
Google97.78%
Healthy24h 98.81%
7d
98.30%
30d
98.30%
p50
393ms
p99
2.01s
Throughput
38 tps
Context
128K
$/Mtok
$0.720 / $0.720
google-vertex
SambaNova96.71%
Degraded24h 96.75%
7d
95.41%
30d
95.41%
p50
672ms
p99
5.68s
Throughput
79 tps◆ (best)
Context
131K
$/Mtok
$0.450 / $0.900
sambanova-turbo
DeepInfra96.02%
Degraded24h 96.82%
7d
96.49%
30d
96.49%
p50
841ms
p99
7.50s
Throughput
7.0 tps
Context
131K
$/Mtok
$0.100 / $0.320◆ (best)
deepinfra/turbofp8
Nebius—
Down24h 41.46%
7d
42.69%
30d
42.69%
p50
1.86s
p99
14.3s
Throughput
7.0 tps
Context
131K
$/Mtok
$0.130 / $0.400
nebius/fp8fp8
Together—
No data24h 98.50%
7d
97.76%
30d
97.76%
p50
680ms
p99
3.14s
Throughput
30 tps
Context
131K
$/Mtok
$1.04 / $1.04
together

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.