Start typing to search.
Ranks providers by estimated time to return 500 tokens, combining the wait for the first token with output speed.
Weights & Biases currently ranks first at 4.58 s via wandb/bf16.
| Rank | Provider / route | Estimated response time | Speed measurements | Endpoint details |
|---|---|---|---|---|
| #1 | Provider / route Weights & Biases wandb/bf16 | Estimated response time4.58 s | Speed measurements
| Endpoint details
|
| #2 | Provider / route Groq groq | Estimated response time5.29 s | Speed measurements
| Endpoint details
|
| #3 | Provider / route NovitaAI novita/fp8 | Estimated response time7.92 s | Speed measurements
| Endpoint details
|
| #4 | Provider / route DeepInfra deepinfra/fp8 | Estimated response time15.18 s | Speed measurements
| Endpoint details
|
| #5 | Provider / route DeepInfra deepinfra/bf16 | Estimated response time23.06 s | Speed measurements
| Endpoint details
|
| #6 | Provider / route Cloudflare cloudflare/fp8 | Estimated response time29.93 s | Speed measurements
| Endpoint details
|
Endpoint data fetched .
Llama 3.1 8B Instruct endpoint guide
6 of 6 Llama 3.1 8B Instruct endpoints currently have the published data required for this ranking. Weights & Biases leads at 4.58 s via wandb/bf16. The table keeps unranked routes visible so missing measurements do not look like missing provider availability.
The fastest ranking estimates a 500-token answer by adding the recent median first-token wait to 500 divided by median output throughput. It represents an example complete response, not a guarantee for every prompt or request size.
Groq currently ranks second at 5.29 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 3.1 8B Instruct has 6 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.
