Last updated: 2026-07-01
Kimi is Moonshot AI's open source multimodal assistant, built on a 1-trillion-parameter Mixture-of-Experts model with 32 billion parameters active per query. Its Agent Swarm feature coordinates sub-agents in parallel across research, coding, and analysis tasks, and the whole stack runs under a Modified MIT license for self-hosting.
About Kimi
Kimi is a frontier AI assistant from Moonshot AI, a Beijing-based lab founded in 2023 that has raised $1.77B, including a $500M Series C at a $4.3B valuation in January 2026. The current Kimi K2.5 model uses a Mixture-of-Experts architecture with 1 trillion total parameters (32 billion active per query), and the weights have been published under a Modified MIT open-source license since July 2025. Both a consumer chat app and a developer API are available; the API is billed per token while the chat app runs on subscription tiers. Kimi is positioned for developers, researchers, and cost-sensitive teams that want frontier-class capability without proprietary lock-in.
Pricing
Free tier (Adagio) with unlimited basic chat but limited agent usage. Consumer subscriptions: Andante $19/month (¥49 CNY), Moderato/Allegretto $39-79/month (¥99-199 CNY), Vivace $159-199/month. API pricing: $0.60 per million input tokens, $2.50 per million output tokens; cached input tokens $0.15/M (75% discount). Turbo model: $1.15/$8.00/M. Automatic $5 voucher when cumulative recharge reaches $5.
Key Features
- Kimi K3 (2.8T Parameters): Kimi K3 launched July 16, 2026 as Moonshot AI's largest open-source model at 2.8 trillion parameters with a 1M-token context window. Moonshot says it performs competitively with Claude Fable 5 and beats Claude Opus 4.8 and GPT-5.6 Sol on its benchmarks, with full open weights due by July 27, 2026 under a Modified MIT licence.
- Kimi K2.6 with Office Pilot: K2.6 (April 2026) introduces K2.6 Thinking for step-by-step math and logic, Office Pilot for lossless conversion across PPT/Word/Excel/PDF with auto-generated formulas, and support for analyzing 50 files at once.
- Native Multimodality: Pre-trained on vision and language tokens with advanced visual understanding, enabling code generation from UI designs and visual task comprehension
- Agent Swarm (300 Sub-Agents): Spawns up to 300 sub-agents in parallel with 4,000+ coordinated steps, cutting execution time by up to 4.5x for complex multi-step research, coding, and analysis tasks.
- Extended Context Window: Supports 256K token context window enabling processing of entire documents, books, and extended conversations without losing context
- Thinking & Instant Modes: Offers both deep reasoning mode for complex problems and instant mode for quick responses, with configurable reasoning depth
- OpenAI-Compatible API: Drop-in replacement for OpenAI API with automatic context caching that lowers costs on repeated content
- 300-Step Tool Calling: Supports autonomous execution of up to 300 sequential tool calls for complex workflows including web search, file operations, and system commands
Pros
- Reviewers consistently cite the price-to-benchmark ratio as the main draw: frontier-class coding and reasoning scores at a fraction of what proprietary APIs charge per token.
- Being open-weight under a Modified MIT license means teams can self-host or fine-tune Kimi instead of depending on a single vendor's uptime and terms.
- Native multimodal input and agentic tool-calling remove the need to bolt on a separate vision model or orchestration framework for common visual-coding and multi-step workflows.
Cons
- Chinese-first interface and primary focus on the Chinese market may limit international support and documentation.
- Younger vendor with less established enterprise support compared to OpenAI and Anthropic.
- Data residency concerns for users requiring non-China infrastructure, since Moonshot AI is based in Beijing.
- Limited established integrations compared to competitor platforms with a longer market presence.
Frequently Asked Questions
How much does Kimi cost in 2026?
Kimi's consumer chat app is free on the Adagio plan; paid tiers run from Andante at $19/month up to Vivace at $159-199/month for higher agent usage. API access is billed separately at $0.60 per million input tokens and $2.50 per million output tokens, with cached input tokens at $0.15/million (a 75% discount) and a faster Turbo endpoint at $1.15/$8.00 per million.
Is Kimi free to use?
Yes, the Adagio tier includes unlimited basic chat at no cost. Agent Swarm and higher-volume agent features are reserved for paid consumer tiers starting at Allegretto, and the developer API carries no free quota; every API call is billed per token from the first request.
What are the best alternatives to Kimi?
ChatGPT is the closest mainstream alternative, with a free plan and a paid tier around $20/month plus a much larger existing user base. DeepSeek is the more direct match: also open source and priced for cost-sensitive API use. Qwen from Alibaba is worth a look for its longer context windows and broader language coverage if you work outside English.
How does Kimi compare to ChatGPT in 2026?
Kimi undercuts ChatGPT on per-token API pricing and is available as open-weight software that teams can self-host; ChatGPT stays closed-source but has a far larger installed user base and broader third-party integrations. Pick Kimi for cost-sensitive, self-hostable deployments; pick ChatGPT for out-of-the-box breadth of integrations.
How do you get started with Kimi?
Sign up for the free Adagio plan at kimi.com to try the chat interface with no cost. For API access, create an account on the Moonshot AI platform, generate an API key, and call the OpenAI-compatible endpoint: most existing OpenAI SDK code only needs a base-URL and key change to start working.
Top Alternatives
- ChatGPT: Pick Kimi if you want a lower per-token API price and open weights; pick ChatGPT for its 900 million weekly users and deeper third-party integrations.
- DeepSeek: Pick DeepSeek if you want another open-source, cost-effective model built around reasoning; pick Kimi if you want native multimodal input and Agent Swarm's parallel task execution.
- Qwen: Pick Qwen for its larger 1M-token context window and 119+ language support; pick Kimi for Agent Swarm's parallel task execution and its MoonViT vision encoder.