Moonshot AI · Long tasks at a lower cost
Kimi K3
A strong model for long analytical tasks and professional outputs. Pricing is more favorable than for some leading models, but completion can take a long time.
- Costs
- Medium to high · approximately $0.94 per Intelligence Index task
- Speed
- Slow on long, multi-step tasks
- Availability
- Moonshot AI's online service
- Input length
- Check the input length for the specific service and version
- Inputs
- Primarily text; depends on the service used
Data checked . Release date unverified.
Indicative capability profile
Where the model is strong
The five levels are our clear summary of the results below. They are not a ranking that applies to every task.
Ideal use
When to choose it
- Long research and knowledge tasks
- A cheaper alternative to the most expensive models
- Background tasks where waiting is acceptable
Usage boundaries
When to choose another model
- Real-time interactive communication
- Short tasks where overhead outweighs the benefit
Traceable supporting sources
Data and measurement sources
Distinguish between the manufacturer's documentation and the results of a specific test. Measurements also depend on the settings and the task set used.
Intelligence Index v4.1
57 pointsThird place among the leading models tracked as of July 17, 2026.
Open original source ↗GDPval-AA v2 · AA-Briefcase
1668 · 1547 Elo pointsA higher number is better. The result confirms strong autonomous work with information.
Open original source ↗Trends over time
Related events from AI Radar
Kimi K3 from Moonshot AI is newly available on the Amazon Bedrock platform
New: Kimi K3 is now available on Amazon Bedrock as a new distribution channel; The model supports prompt caching on Bedrock; Data in AWS Bedrock is protected by the AWS data boundary without being shared with the provider; The Bedrock deployment optimizes the model for long coding and knowledge workflows
Developers can try Kimi K3 through the API or through tools such as OpenCode for agentic programming tasks with long context, but must account for a higher price per token than with previous Kimi models.
Companies can deploy Kimi K3 through Amazon Bedrock with data protection from AWS without needing their own GPU infrastructure, or host the model themselves on instances with 8 NVIDIA B300 GPUs – but both options involve higher costs than earlier open Kimi models.
Moonshot AI releases the Kimi K3.1 Pro Highspeed model on the Amazon Bedrock platform
Developers who already call Kimi K3 via Amazon Bedrock can switch to K3.1 Pro Highspeed without changing the API, just by adjusting the model ID, and can influence speed/price using the reasoning_effort parameter or by choosing the Flex tier.
Companies building agentic systems or coding assistants on Amazon Bedrock get a faster and, according to Moonshot AI, more accurate model with preserved data isolation (zero data retention, no access by AWS operators, no data sharing with Moonshot AI) and a choice of regional data residency (US/India), which is relevant for compliance requirements.
Kimi K3 model from Moonshot AI available on Amazon Bedrock with inference in India
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.
Moonshot AI is in talks with Microsoft, Amazon and Google about hosting the Kimi K3 model in exchange for a share of revenue
If the talks result in an agreement, developers could eventually access the Kimi K3 model through the interfaces of familiar cloud platforms instead of connecting directly to a Chinese provider, but this remains uncertain for now.
If an agreement were reached, companies could deploy the Kimi K3 model through standard Azure, AWS or Google Cloud infrastructure instead of connecting directly to a Chinese provider, making compliance and procurement easier; at the same time, this is a deal with heightened political risk due to the threat of a trade embargo and ongoing accusations against Moonshot AI.
Next step
Compare the model using your actual task.
A benchmark narrows the selection. A short trial on your data determines the choice.