← all entries
C-TIER news

Kimi K3 — hyped open-weight 2.8T model claims to beat Fable 5 cheaply

tenfoldmarc · 2026-08-14T22:20:18.000Z · in AI Saves, Vol. 8
Open post ↗

100k ft — What is this?

Kimi K3 — hyped open-weight 2.8T model claims to beat Fable 5 cheaply
newsC-tier

50k ft — Why care?

Hype clip claiming Moonshot's Kimi K3 (2.8T parameters, open source) beat Claude Fable 5 on coding and design at ~5x lower cost, rattled the Nasdaq, and built a CS:GO portal clone for ~$3 in tokens. No benchmarks shown; link gated behind a comment.
LLMdesignnewsopen sourcesoftware development

10k ft — Key takeaways

• Market-moving claims here are unsourced. • Price-performance pressure from open-weight models is real; verify claims.

5k ft — Tools & techniques

Anthropic, Claude, Kimi, OpenAI API integration, code generation

1k ft — Deep-dive / try it

This new model is putting pressure on OpenAI and Anthropic. Comment "Kimi" and I'll send you a link to try it out. #kimiai #ai #artificialintelligence
Transcript excerpt
On the day that Kimi K3 dropped, the NASDAQ fell 1% and Nvidia got dumped on. So how does a free open source model become Wall Street's worst nightmare? Well, I didn't think this was possible but Kimi just destroyed Fable of Design while being five times cheaper. Even Elon Musk said, this is the standard to beat. Now this is Kimi K3, the largest open source model in history with 2.8 trillion parameters. It beat Fable 5 at coding and design but cost about five times less. And unlike Claude or GPT, you can actually use this for free. And people are calling this another deep seek moment and it wiped out the entire US stock market after the second day was released. This is how someone built this CSGO portal clone with just over $3, yes $3 in API tokens. But what's even better is in the design test that I've seen done, it beats Fable 5 which I never expected. So if you want access to try it out, just comment Kimi and I'll send you the full setup plus the link to try it out yourself.

More from @tenfoldmarc

Collection: AI Saves, Vol. 8