Same open weights, served by 2 providers. Weights: meta-llama/Llama-3.1-8B-Instruct.
| Provider | Their model id | Serving | Input | Output | Cached in | Blended 3:1 | Source | |
|---|---|---|---|---|---|---|---|---|
DeepInfra | meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo | standardFP8 | $0.02 | $0.04 | — | $0.025 | Verified 4h ago · Changed 7h ago | |
DeepInfra Deprecated, 2026-07-16 | meta-llama/Meta-Llama-3.1-8B-Instruct | standardFull precision | $0.02 | $0.05 | — | $0.028 | Verified 4h ago · Changed 7h ago | |
AWS Bedrock | us.meta.llama3-1-8b-instruct-v1:0 | standardQuantization not disclosedReduced context (128k of 131k) | $0.22 | $0.22 | — | $0.22 | Verified 6h ago · Changed 6h ago | |
AWS Bedrock | meta.llama3-1-8b-instruct-v1:0 | standardQuantization not disclosedReduced context (128k of 131k) | $0.22 | $0.22 | — | $0.22 | Verified 6h ago · Changed 6h ago |
Same weights don't always mean identical outputs: quantization, context limits and serving stacks can change quality, especially for tool calling. Prices are USD per 1M tokens; hyperscaler prices are for the reference region shown.
| DeepInfra fp8 | $2.8/mo |
| DeepInfra bfloat16 | $3/mo |
| AWS Bedrock | $26.4/mo |
| AWS Bedrock | $26.4/mo |
Cache hits use each provider's cached-input price where it has one. Cache write costs aren't included.