Umans Kimi K3 Deprecated
deprecated · serving until Nov 9, 2026 · successor umans-kimi-k3.1-pro
66.1tok/s
throughput · p50 · last 5 min
1.32s
TTFT · p50 · last 5 min
99.65%
uptime · 24h
Kimi K3: Moonshot's most capable model and the first open 3T-class release - 2.8T parameters, a 1M-token context window, and native vision, built for repository-scale understanding and long agentic runs. On Artificial Analysis' independent Intelligence Index it scores 44 to Claude Fable 5's 50 at about a third of the cost to run it, and it leads Fable 5 on AutomationBench-AA (58% vs 54%). It thinks by default at maximum reasoning effort; select none, low, high, or max to trade depth for speed. Deprecated: use `umans-kimi-k3.1-pro` instead (sunset 2026-11-09).
Context
1049K
Max output
131K
Recommended
131K
Vision
Yes
Tools
Yes
Reasoning
Toggle · none/low/high/max
Weights
Trends
Speed over the last 90 days
90 days agopre-release before Jul 31, 2026today
90 days agopre-release before Jul 31, 2026today
Changelog
Events for Umans Kimi K3
Nov 62026
Deprecated: Umans Kimi K3, retiring November 9
umans-kimi-k3 is deprecated and retires on Monday, November 9, 2026. Its successor is umans-kimi-k3.1-pro: Kimi K3.1 Pro, Moonshot's newest open model, post-trained from Kimi K3, with the same 1M-token context, native image understanding and reasoning settings (thinks at max effort by default), at a lower price: $2.00 / $10.00 / $0.20 per 1M (input / output / cache read) against $3.00 / $15.00 / $0.30. Switch your model id to umans-kimi-k3.1-pro before November 9. In the Umans CLI, Claude Code's fable tier uses Kimi K3.1 Pro from version 0.1.82.
Older events 2
Aug 12026
Generally available: Umans Kimi K3 Released
The Kimi K3 Labs pilot closed and umans-kimi-k3 is generally available: no Labs seat required anymore: plan and pay-per-token keys alike get the 1M context window, native vision, and max-effort reasoning at $3.00 / $15.00 / $0.30 per 1M (input / output / cache read).
Jul 312026
Released pay-per-token: Umans Kimi K3 Released
umans-kimi-k3 opened to pay-per-token (wallet and service-account keys): Moonshot's largest open model, a 1M context window, native vision, and max reasoning effort by default, billed at $3.00 / $15.00 / $0.30 per 1M (input / output / cache read).