Kimi K3 OpenRouter Prices Shift as AWS Adds India Inference
Esta notícia ainda só está disponível em inglês.
MoonshotAI's Kimi K3 saw a mixed price adjustment on OpenRouter on November 4, 2026, while AWS Machine Learning announced in-country inference support for the model on Amazon Bedrock. The 2,779.9B parameter open-weight model now costs $0.276 per million input tokens and $9.60 per million output tokens, reflecting a 43.8 percent increase in input costs and a 36.0 percent decrease in output costs compared to November 3, 2026.
OpenRouter Price Changes
The input price for Kimi K3 rose from $0.192 to $0.276 per million tokens. Simultaneously, the output price fell from $15.00 to $9.60 per million tokens. The model maintains a context window of 1,048,576 tokens on the platform.
These changes mark the first price update for the model since its release on July 16, 2026. The divergence between input and output pricing adjustments results in a varied cost structure for developers using the API.
AWS Bedrock Availability
AWS Machine Learning announced that Amazon Bedrock now supports in-country inference in India for Kimi K3. According to AWS Machine Learning, the model is offered through cross-Region inference profiles, with the global profile ID set as global.moonshotai.kimi-k3 and the India geographic profile ID as in.moonshotai.kimi-k3.
AWS Machine Learning states that India geographic inference processing must stay within Mumbai and Hyderabad Regions. The service supports images, PDFs, and other file types, and includes a flex tier with a 50 percent discount on default standard pricing. Additionally, the platform offers discounted pricing for cache hits and an explicit caching mode.
According to AWS Machine Learning, Ultrahuman cut LLM costs by 80 percent with prompt caching and made multi-turn chats 85 percent faster. Amazon Bedrock Advanced Prompt Optimization is available as a fully managed option to optimize prompt templates.
Pontos principais
- Kimi K3 input prices on OpenRouter increased by 43.8 percent while output prices decreased by 36.0 percent.
- AWS Machine Learning enabled in-country inference for Kimi K3 in India via Amazon Bedrock.
- The model supports a 1,048,576 token context window and is available as an open-weight foundation model.
Anúncio oficial
Publicado por AWS Machine Learning Blog. Leia na fonte.
Introducing India-based inference for Kimi K3 on Amazon Bedrock