Kimi K3 matches Claude Opus 4.8 on measured intelligence, wins entire categories outright, and ships its weights on July 27. The full guide: where it wins, where it lies, and the exact setup to run it
On Thursday, a Beijing lab crossed a line the industry assumed was years away.
Moonshot AI shipped Kimi K3: 2.8 trillion parameters, a 1M-token context window, native vision, and the promise of full open weights by July 27.
The independent verdict landed within 48 hours, and it reads like a turning point:
▫️ 57 on the Artificial Analysis Intelligence Index. Claude Opus 4.8 scores 56. Fable 5 and GPT-5.6 Sol stay ahead, and everything else sits behind an open-weight model from China.
▫️ #1 outright on AutomationBench (Zapier-style SaaS agent workflows, 53%), #1 on Arena’s Frontend Code leaderboard ahead of Fable 5, and the best published BrowseComp score anywhere at 91.2%.
▫️ $0.94 per completed task on the index, half of Opus 4.8’s $1.80.
The market understood immediately. Zhipu fell 28.4% the next day. MiniMax dropped 15.6%. Bank of America wrote that Moonshot proved you can reach the frontier on restricted chips by out-designing the training, and Simon Willison had it running the same morning.
Here is the free insight, and it is the one that reframes the launch: this ends the era where “open-weight” meant “the cheap tier.” K3 prices at $3 per million input and $15 per million output, Sonnet 5’s exact rate card. Moonshot is charging Western mid-tier prices because the measured intelligence justifies it, and in eight days the market got Grok 4.5, GPT-5.6, and K3, three different labs attacking the frontier from three different directions. Six labs now score above 50 on the index. In early June it was two.
Which means the question changed again. It went from “which model is smartest” to “which model wins this lane at this price,” and K3 just took entire lanes. It also ships with two genuine problems the launch coverage skipped, one of them a number that should give every builder before routing production traffic to it.
Behind the paywall, The Kimi K3 Playbook:
▫️
independent numbers only, with the self-reported claims separated from what testing confirmedThe verified scoreboard,▫️
frontend code, browsing agents, SaaS automations, long-horizon knowledge work, and the surprise fifthThe five lanes K3 wins,▫️
the hallucination number Moonshot skipped mentioning, the verbosity tax, and the price-card trapThe honest file,▫️
per-task figures against Fable 5, Opus 4.8, Grok 4.5, and the open-weight peers, plus the two demos where the gap hits 12xThe cost math,▫️
API, OpenRouter, and the Kimi Code CLI, with the exact configThe 48-hour setup,▫️
what the July 27 weights actually require, and who should skip themThe self-host reality check,▫️
exactly where K3 slots into the model-routing table, lane by laneThe router update,▫️
the five events that change this playbookThe re-route triggers,
One subscription unlocks every system
This is one build in a growing library. Premium opens all of them:
▫️ [The AI Tools and Models library](https://www.the-ai-corner.com/t/ai-tools-and-models?r=1krivi)
▫️ [The Business and Investing library](https://www.the-ai-corner.com/t/business-and-investing?r=1krivi)
Plus 3 fresh systems every week. One correctly routed workload pays the subscription back before the weights even ship.
The verified scoreboard, the five lanes, the honest file, the cost math, the 48-hour setup, and the router update, in one system.
Get The Kimi K3 Playbook below 👇
Try premium free for 7 days. Or get 50% off this week only:
Keep reading with a 7-day free trial #
Subscribe to The AI Corner to keep reading this post and get 7 days of free access to the full post archives.