Together introduces US-processed serverless inference endpoints
一句重點：The first seven open-source models can use an endpoint that keeps inference in the US; switching requires only a base URL change and costs 25% more.
發生什麼事
@togethercompute announced region-based serverless inference endpoints. Its initial group covers seven leading open-source models, with inference handled entirely in the US. To switch, users only need to replace the base URL; their API key and model IDs remain unchanged, while pricing is 1.25 times the standard rate. The post does not list the seven models or mention plans for other regions (🗣 self-reported).
01Sources
@togethercompute: Together’s official account announcing serverless inference endpoints processed in the US