Start typing to search.
Ranks providers by the uptime OpenRouter measured during the latest 30-minute window.
Cloudflare currently ranks first at 100.00% via cloudflare/fp8.
| Rank | Provider / route | Uptime (last 30 min) | Speed measurements | Endpoint details |
|---|---|---|---|---|
| #1 | Provider / route Cloudflare cloudflare/fp8 | Uptime (last 30 min)100.00% | Speed measurements
| Endpoint details
|
| #2 | Provider / route Weights & Biases wandb/bf16 | Uptime (last 30 min)100.00% | Speed measurements
| Endpoint details
|
| #3 | Provider / route Groq groq | Uptime (last 30 min)100.00% | Speed measurements
| Endpoint details
|
| #4 | Provider / route NovitaAI novita/fp8 | Uptime (last 30 min)100.00% | Speed measurements
| Endpoint details
|
| #5 | Provider / route DeepInfra deepinfra/fp8 | Uptime (last 30 min)99.98% | Speed measurements
| Endpoint details
|
| #6 | Provider / route DeepInfra deepinfra/bf16 | Uptime (last 30 min)99.81% | 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. Cloudflare leads at 100.00% via cloudflare/fp8. The table keeps unranked routes visible so missing measurements do not look like missing provider availability.
The reliability ranking uses the published 30-minute endpoint uptime measurement. This short operational window can reveal current instability, but it is not a service-level agreement or a substitute for longer-term availability monitoring.
Weights & Biases currently ranks second at 100.00%. 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.
