Start typing to search.
Ranks providers by estimated cost using a 1,000-input/500-output token ratio, shown per 1 million total tokens.
DeepInfra currently ranks first at $0.023333 / 1M tokens via deepinfra/fp8.
| Rank | Provider / route | Estimated cost / 1M tokens | Speed measurements | Endpoint details |
|---|---|---|---|---|
| #1 | Provider / route DeepInfra deepinfra/fp8 | Estimated cost / 1M tokens$0.023333 / 1M tokens | Speed measurements
| Endpoint details
|
| #2 | Provider / route NovitaAI novita/fp8 | Estimated cost / 1M tokens$0.03 / 1M tokens | Speed measurements
| Endpoint details
|
| #3 | Provider / route DeepInfra deepinfra/bf16 | Estimated cost / 1M tokens$0.03 / 1M tokens | Speed measurements
| Endpoint details
|
| #4 | Provider / route Groq groq | Estimated cost / 1M tokens$0.06 / 1M tokens | Speed measurements
| Endpoint details
|
| #5 | Provider / route Cloudflare cloudflare/fp8 | Estimated cost / 1M tokens$0.197 / 1M tokens | Speed measurements
| Endpoint details
|
| #6 | Provider / route Weights & Biases wandb/bf16 | Estimated cost / 1M tokens$0.22 / 1M tokens | 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. DeepInfra leads at $0.023333 / 1M tokens via deepinfra/fp8. The table keeps unranked routes visible so missing measurements do not look like missing provider availability.
The cheapest ranking uses a 1,000-input/500-output-token ratio and reports the blended result per 1 million total tokens. Because providers can price input and output differently, a workload with a different token ratio may produce a different order.
NovitaAI currently ranks second at $0.03 / 1M tokens. 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.
