Start typing to search.
Ranks providers by their recent typical output speed in tokens per second. Providers without enough recent data remain visible after ranked rows.
Groq currently ranks first at 145.0 tok/s via groq.
| Rank | Provider / route | Output speed | Speed measurements | Endpoint details |
|---|---|---|---|---|
| #1 | Provider / route Groq groq | Output speed145.0 tok/s | Speed measurements
| Endpoint details
|
| #2 | Provider / route Weights & Biases wandb/fp16 | Output speed54.0 tok/s | Speed measurements
| Endpoint details
|
| #3 | Provider / route SambaNova sambanova-turbo | Output speed52.5 tok/s | Speed measurements
| Endpoint details
|
| #4 | Provider / route Google Vertex google-vertex | Output speed49.0 tok/s | Speed measurements
| Endpoint details
|
| #5 | Provider / route Parasail parasail/fp8 | Output speed45.0 tok/s | Speed measurements
| Endpoint details
|
| #6 | Provider / route Cloudflare cloudflare/fp8 | Output speed29.0 tok/s | Speed measurements
| Endpoint details
|
| #7 | Provider / route AkashML akashml/fp8 | Output speed27.0 tok/s | Speed measurements
| Endpoint details
|
| #8 | Provider / route Nebius nebius/fp8 | Output speed25.0 tok/s | Speed measurements
| Endpoint details
|
| #9 | Provider / route Together together/fp8 | Output speed24.0 tok/s | Speed measurements
| Endpoint details
|
| #10 | Provider / route Google Vertex google-vertex/us-central1 | Output speed16.0 tok/s | Speed measurements
| Endpoint details
|
| #11 | Provider / route DeepInfra deepinfra/turbo | Output speed7.0 tok/s | Speed measurements
| Endpoint details
|
| #12 | Provider / route NovitaAI novita/bf16 | Output speed5.0 tok/s | Speed measurements
| Endpoint details
|
Endpoint data fetched .
Llama 3.3 70B Instruct endpoint guide
12 of 12 Llama 3.3 70B Instruct endpoints currently have the published data required for this ranking. Groq leads at 145.0 tok/s via groq. The table keeps unranked routes visible so missing measurements do not look like missing provider availability.
The output-speed ranking compares recent median generated tokens per second. It favors routes that complete long generations quickly, while first-token latency remains a separate measure of how responsive the endpoint feels at the beginning.
Weights & Biases currently ranks second at 54.0 tok/s. Compare that gap with input and output price, response-start time, output speed, 30-minute uptime, context length, and the exact route before deciding whether first place is meaningful for your workload.
Llama 3.3 70B Instruct has 12 published provider options, and performance data is matched to each exact OpenRouter routing tag. Different quantization, context, regional deployment, or provider configuration can change price and behavior even when the underlying model name is identical.
