For anyone who wants the methodology: same prompt, same rubric locked before either plan was scored + a full breakdown of where the two models agreed and split. blog.kilo.ai/p/glm-52-vs-kim…

$1.40 in / $4.40 out vs $10 / $50. That's GLM-5.2 against Claude Fable 5, per million tokens. On the same planning task, scored on the same rubric, they came out 9.0 and 9.1. Near-identical plans. A tenth of the price. What would you build differently if frontier-level…

And then there's Le Chaton Fat. Crushed every benchmark we ran. We have no notes.

Grok Build 0.1 beat frontier models on our Code Reviews benchmark. We wrote up what we saw:

US export controls forced Anthropic to suspend Fable 5 last week. We benchmarked GLM-5.2 and Kimi K2.7 Code head-to-head. The buried lede: GLM scored 9.0 against Fable's 9.1, at a tenth of the price. Open-weight is closing the gap. Full writeup from @TheBinaryNeuron:…