Official facts before secondary claims
Use this page when you need the public Kimi and Moonshot pages that can support Kimi K3 claims without mixing in third-party commentary.
This short video summarizes the page: the source set, the decision boundary and the practical next action.
The Kimi K3 blog is the primary source for the 2.8T-parameter open 3T-class positioning, native vision, 1M-token context, KDA, Attention Residuals and the pending technical report and full-weight release notes.
The help center and model-mode pages are better for product behavior: K2.6 versus K3 versus K3 Cluster, multimodal input support, tool workflows, credit usage and membership context limits.
Official sources can support direct Kimi K3 facts. External reviews, community screenshots and third-party measurements can support interpretation only when they are clearly labeled.
Keep official facts, verified docs and pending items distinct when preparing a Kimi K3 page or brief.
Cards are generated from the clauxel/kimik3 source ledger and keep source type, priority and status visible.
Official launch page for Kimi K3, including the 2.8T-parameter framing, KDA, Attention Residuals, native vision, 1M-token context and pending report/weight notes.
Moonshot company page that mirrors the K3 positioning in Chinese and routes readers to Kimi, API, Kimi Code and product lines.
Official help entry for Kimi model selection, multimodal inputs, files, URLs, videos and output formats.
Official product timeline showing Kimi K3 as the capability base for Chat, Agent, Agent Swarm, Kimi Work, Kimi Code and API.
Official mode selection guide that separates K2.6, K3 and K3 Cluster fit, reasoning intensity and membership context limits.
Official Kimi Websites help page connecting K3 vision and coding ability to website generation and iterative edits.