Amazon Web Services (AWS) has introduced Kimi K3 India inference on Amazon Bedrock, letting enterprises run Moonshot AI’s open-weight Kimi K3 model with all processing kept inside its Mumbai and Hyderabad regions. AWS announced the option on 3 November 2026 for workloads with strict data residency rules, such as banking and public services.

The launch adds a third choice to the global and US-based inference already offered for the model on Bedrock. It also widens the menu of in-country AI options for Indian firms, alongside home-grown services such as E2E Networks’ pay-per-token Blaze inference platform.

How Kimi K3 India inference works

Kimi K3 is a multimodal, open-weight foundation model from Moonshot AI. On Bedrock it is served through cross-Region inference profiles, where the profile ID a developer calls decides which AWS regions may process the request, according to the AWS Machine Learning Blog announcement.

• Global profile (global.moonshotai.kimi-k3): recommended by AWS for most uses as the lowest-cost and most widely available option.
• India profile (in.moonshotai.kimi-k3): keeps inference within the Mumbai (ap-south-1) and Hyderabad (ap-south-2) regions.
• Network path: traffic between regions is encrypted in transit and stays on the AWS network rather than the public internet.

AWS added one caveat for compliance teams. Logs and metrics are stored in the region where a request is sent, not where it is processed, so a team calling from outside India would keep those records abroad. Organisations that restrict regions must also allow both Indian destinations in their access policies.

Built for Indic documents and Hinglish

AWS demonstrated the model reading a scanned Hindi electricity bill, extracting fields such as the amount due and due date as structured data, and answering follow-up questions in Hinglish, without a separate OCR or translation step. It pointed to KYC forms and utility bills as common use cases in Indian financial services and the public sector.

The model can be called through OpenAI-compatible Chat Completions and Responses APIs, as well as Bedrock’s Converse and InvokeModel APIs, with support for streaming, tool calling and structured outputs. AWS also noted a flex service tier priced at a 50% discount to standard rates for less time-sensitive jobs.

Ultrahuman is already using it

Wearables maker Ultrahuman runs Kimi K3 on Bedrock in Jade, the AI health and wellness assistant in its app, which handles more than 14,000 conversations a day using each user’s ring and glucose data. Aditya Patawari, VP Engineering at Ultrahuman, said in the AWS post that prompt caching cut the company’s LLM costs by 80% and made multi-turn chats 85% faster. These are the company’s own figures.

Why it matters for Indian enterprise AI

Data residency has been one of the main hurdles for regulated Indian sectors that want frontier-class models. In-country processing for a leading open-weight model fits the wider sovereign AI push seen in Mistral’s open-weight Le Chonk release and its sovereign AI playbook for India.

It also follows a build-out of local compute, from STTGDC’s 22 MW Noida data centre hosting Neysa AI GPUs to global AI companies turning to India for data centre capacity.

Key takeaways

• AWS now offers India-only Kimi K3 inference on Amazon Bedrock through the in.moonshotai.kimi-k3 profile.
• Processing stays in the Mumbai and Hyderabad regions, but logs follow the region the request is sent from.
• Ultrahuman reports 80% lower LLM costs with prompt caching in its Jade assistant.

FAQ

Which AWS regions process Kimi K3 India inference?

Requests made with the India profile are processed only in the AWS Mumbai (ap-south-1) and Hyderabad (ap-south-2) regions.

Who is it meant for?

AWS said it targets workloads with strict data residency needs, such as Indian financial services and public sector document processing.

Featured image: Welcome entrance at AWS Summit Mumbai 2026. Photo: Zaidusyy, Wikimedia Commons (CC BY-SA 3.0)

Be a part of Elets Collaborative Initiatives. Join Us for Upcoming Events and explore business opportunities. Like us on Facebook , connect with us on LinkedIn and follow us on Twitter.

"Exciting news! Elets technomedia is now on WhatsApp Channels Subscribe today by clicking the link and stay updated with the latest insights!" Click here!