Skip to main content
Moonshot AI/Pro+

Kimi K3

Moonshot AI's 2.8-trillion-parameter flagship — the world's first open 3T-class model. Built for sustained, multi-step coding and agent workflows with a 1M token context window and native multimodal input.

1M tokens
Context Window
Pro+
Minimum Plan
Moonshot AI
Provider

Best for

Long-horizon coding sessions, frontend/UI generation, multi-step agent workflows, large codebase analysis

Strengths

2.8T parameters (MoE, 16/896 experts active per token)

#1 on Frontend Code Arena (blind human evaluation)

Excels at sustained multi-step engineering tasks

Native vision — debug UI against screenshots

1M token context with flat pricing, no long-context surcharge

Always-on chain-of-thought reasoning

When to use

Choose Kimi K3 when you need sustained focus across large codebases, multi-file refactoring, frontend code generation, or agent workflows that require many sequential tool calls. Particularly strong when the task demands endurance over multiple iterations rather than a single-shot answer.

Limitations

Always-on reasoning means every response includes a thinking trace (billed as output tokens), making simple queries more expensive than lightweight models. No non-thinking mode available yet.

Try Kimi K3 on Multos AI

42 models available. Switch per message. Start free.

Get Started Free