A post-trained successor to K3 with sparse attention, positioned as Moonshot’s strongest open-weight agent and coding model. It has a 1M tokens context window, and its weights are openly available. Check Moonshot AI’s pricing page for current rates.
Among the tools we track, it’s available in Amp. A faster highspeed variant launched alongside it. It’s also part of ClinePass, and Moonshot documents using it with Codex and Claude Code through its own API.
What Moonshot AI reports
- Moonshot reports 65.7 on Terminal-Bench 4.0
- Moonshot reports 75.5 on DeepSWE v1.1
Vendor-reported figures, not independently verified by Code×Radar.
Related changes
- Cline 4.1.26 changes default models and adds a model setting for cloud tasks · October 26, 2026
Sources
Facts last checked November 15, 2026.