Kimi K2.5

Best Value

Previous-generation flagship with Agent Swarm, 1T MoE, and 128K context.

Agent Swarm128K context

Same provider

Kimi

Flagship model for long-horizon coding and end-to-end knowledge work with 1M context.

Kimi

Coding-focused model with high success rates in long contexts and a 256K window.

Kimi

Same model as K2.7 Code at roughly 180 tokens per second, reaching 260 in short contexts.

Kimi

General-purpose model for stable long-horizon coding, instruction following, and self-correction.

More from Kimi