← The Polylog AI Intelligence Brief
Morning Edition · Tuesday, July 21, 2026Published at 1:32 AM EDT · New York
Another Chinese Open-Weight Model Reaches the Frontier at a Third of the Price
Kimi K3 lists at $3 per million input tokens and $15 per million output tokens, and Ben Thompson argues the pressure falls on marginal cost, not on the top US labs.
Kimi K3, another open-weight model from China, has arrived close to the state of the art and undercuts leading US pricing, according to Ben Thompson's Stratechery analysis. Kimi K3 lists at $3 per million input tokens and $15 per million output tokens, below the $5 and $30 Thompson cites for a comparable US closed model. The release continues a pattern in which downloadable Chinese models reach the frontier within months while charging far less to serve.
Thompson's argument, summarized by Simon Willison, corrects a common confusion. "Open weights" does not mean free to run. The research-and-development spend is a fixed cost, independent of usage, and inference still costs money per token. What open weights change is the competitive floor. When a near-frontier model can be downloaded and served by anyone, price competition moves to marginal cost, and closed labs lose the ability to charge a premium for capability that is now freely available in a comparable open form.
His conclusion runs counter to the reflexive worry about Chinese models. The top US labs, he contends, will be fine because their advantage was never a durable barrier tied to any single model. The real gap is the absence of a strong open US alternative, which leaves governments and companies that are blocked from American frontier models standardizing on Chinese weights by default.
Veracity: Corroborated
85/100- If true, who benefits
Chinese labs and the argument for open-weight distribution, and buyers outside US frontier access, while the pressure falls on Western labs that monetize through metered closed APIs.
- The nuance
Kimi K3's $3 and $15 pricing checks out, but The Decoder notes K3 raised Moonshot's own prices roughly sixfold over K2.6, so the "far cheaper to serve" narrative is weakening even as it still undercuts comparable US closed models.
Part of a tracked trend
Chinese Open-Weight Models Emerge as the Non-US AI Stack
As Washington restricts foreign access to US frontier models, governments and enterprises cut off from American AI increasingly standardize on downloadable Chinese open-weight models, splitting the world into competing AI supply blocs rather than a single frontier.
Townsquare
Start a discussion in Townsquare.
More from this edition
- Google Is Building a Chip That Bakes Gemini's Architecture Into Silicon
- OpenAI Paused an Internal Model After It Found and Exploited a Sandbox Flaw
- An Autonomous AI Agent Breached Hugging Face and Logged 17,000 Actions
- An Anthropic Researcher Says Claude Fable 5 Produced a Counterexample to the Jacobian Conjecture
- Nvidia Puts AI Agents to Work Building Simulation Worlds at SIGGRAPH
- Bristol Myers Squibb Commits to an Nvidia Vera Rubin AI Factory for Drug Discovery
- Meta's Brain2Qwerty Decodes Typed Sentences From Brain Scans at 61% Word Accuracy
- A Paper Finds Open-Weight Models Commit to Answers Before They Reason
Comments
0No comments yet.