July 18, 2026
AI price war, comment section edition
The Kimi K3 Moment
Cheap new AI shocks fans as commenters ask: is this hype, cope, or game over for Claude
TLDR: A developer says Kimi K3 matches Claude for coding while costing far less, turning a model comparison into a bigger fight about pricing, limits, and US policy. Commenters were split between laughing at the hype, demanding clearer tests, and saying the cheaper tool may still waste time on real work.
A spicy new claim lit up the comments: one developer says Kimi K3 feels basically the same as Claude for everyday coding, but at a price that makes Claude look painfully expensive. That alone was enough to spark a mini-meltdown, especially when the post piled on bigger accusations: American AI companies are overcharging, government rules are backfiring, and Chinese-made models are slipping past with fewer limits and lower prices. In plain English, the argument is that users in the US may be getting the most restricted and most expensive options while everyone else gets cheaper tools that just work.
But the comments were where the real fireworks started. One skeptic immediately hit the brakes with the very fair "Which Claude are we even talking about?" question, because saying two tools are identical means a lot less if nobody knows which version was tested. Another commenter tossed in a sarcastic grenade: "Was this written by Kimi K3 or Fable?" That joke basically summed up the thread’s vibe: half fascinated, half suspicious, and fully ready to clown on the post.
Not everyone was buying the miracle sale. One user said Chinese models can "overthink" simple tasks, while another brought receipts with a personal test claiming Kimi chewed through time and usage limits on a real project. Even the boring question about a cost calculator felt revealing: people aren’t just debating quality anymore, they’re trying to survive the pricing drama.
Key Points
- •The article says Kimi K3 and Claude performed similarly on the author’s normal coding tasks, with similar output quality and token usage.
- •It states Kimi K3’s API pricing is $3 per million input tokens and $15 per million output tokens, compared with $10 and $50 for Claude’s top model.
- •The article says Claude’s $20 plan could not sustain Fable access and falls back to Opus, while Kimi’s tiers are described as not having that limitation.
- •It cites Semgrep as finding GLM 5.2 outperforming Claude on cyber benchmarks because the restricted model declined tasks that the open model completed.
- •The article argues that U.S. AI policy and model restrictions are weakening the competitiveness of American AI products relative to Chinese and open alternatives.