Kimi K3 model from Moonshot AI available on Amazon Bedrock with inference in India
Amazon Bedrock offers the Kimi K3 model from Moonshot AI with inference tied to India (Mumbai, Hyderabad) to meet data residency requirements, alongside the existing global profile.
Amazon Bedrock has expanded its offering of the Kimi K3 model from Moonshot AI with the option of geographic inference tied to India. Alongside the existing global profile (global.moonshotai.kimi-k3), which according to the company is the cheapest and most accessible, there is now a new profile, in.moonshotai.kimi-k3, which guarantees that request processing stays exclusively within the Mumbai and Hyderabad regions. According to AWS, the data is encrypted in transit and remains within the AWS network without passing over the public internet.
Kimi K3 is described as a multi-modal model with open weights that, according to Moonshot AI, achieved frontier-level performance in coding and general agentic intelligence at launch. The article cites as an example deployment the Jade app (an AI assistant for health and wellness in the Ultrahuman app), which processes over 14 000 conversations daily based on users' ring and glucose data. According to Ultrahuman's VP Engineering, deploying the Kimi K3 model via Bedrock made it possible to keep health data within AWS with the Mumbai region as the primary, and thanks to prompt caching the company reduced LLM costs by 80% and sped up multi-turn conversations by 85%.
As a specific use case, the article describes OCR and document reading in Indian languages (e.g., Hindi) followed by querying in code-mixed Hinglish, typical for KYC forms and similar documents in financial services and the public sector in India. According to AWS, the model handles this task without a separate OCR or translation service, by directly reading images and PDFs. The rest of the technical details (API, settings) are missing from the source — see the source article for details.
Why it matters
For companies in India or with Indian customers in regulated sectors (finance, healthcare, public sector), geographic inference addresses the requirement that sensitive data never leave the given jurisdiction, without having to operate their own infrastructure. The Ultrahuman example shows a concrete economic benefit of prompt caching (an 80% cost reduction) and speed (an 85% improvement) at high traffic volume (14 000+ conversations daily), which is a relevant figure for cost planning of similar deployments.
Relevant practical impact
What this means
For a business
Companies processing data of Indian users (healthcare, finance, public administration) can deploy the open-weight model Kimi K3 via Bedrock with a guarantee that inference takes place only in India, and thanks to prompt caching significantly reduce LLM operating costs, similar to the Ultrahuman case.
Risks and complianceCheck the original
Event sources
only one source so far · 1 publisher, 0 independent. We count feeds from the same owner only once.