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.
Google Vertex currently ranks first at 151.0 tok/s via google-vertex/us-west2.
| Rank | Provider / route | Output speed | Speed measurements | Endpoint details |
|---|---|---|---|---|
| #1 | Provider / route Google Vertex google-vertex/us-west2 | Output speed151.0 tok/s | Speed measurements
| Endpoint details
|
| #2 | Provider / route SambaNova sambanova/fp8 | Output speed53.0 tok/s | Speed measurements
| Endpoint details
|
| #3 | Provider / route Weights & Biases wandb/fp8 | Output speed47.0 tok/s | Speed measurements
| Endpoint details
|
| #4 | Provider / route MARA mara | Output speed46.5 tok/s | Speed measurements
| Endpoint details
|
| #5 | Provider / route NovitaAI novita/fp8 | Output speed18.0 tok/s | Speed measurements
| Endpoint details
|
| #6 | Provider / route SiliconFlow siliconflow/fp8 | Output speed18.0 tok/s | Speed measurements
| Endpoint details
|
| #7 | Provider / route DeepInfra deepinfra/fp4 | Output speed9.0 tok/s | Speed measurements
| Endpoint details
|
| #8 | Provider / route AtlasCloud atlas-cloud/fp8 | Output speed4.0 tok/s | Speed measurements
| Endpoint details
|
Endpoint data fetched .
DeepSeek V3.1 endpoint guide
8 of 8 DeepSeek V3.1 endpoints currently have the published data required for this ranking. Google Vertex leads at 151.0 tok/s via google-vertex/us-west2. 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.
SambaNova currently ranks second at 53.0 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.
DeepSeek V3.1 has 8 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.
