💺
@patience_cave
patience, jimmy, openai is nothing without its good research takes time in the patience cave.
884 Following 7.5K Followers
MazeBench vs Fable 5.2 vs Gemini 4 Carbon
Fable 5.2 scores 15%, just a hair beneath Astra for half the price. After 40 hours, it is by far the most token efficient model.
Gemini 4 Carbon scores 11%, and lived in the shadow of its predecessor Gemini 4 Argon throughout the run.
Show more
another huge week ahead, for waiting!
announcing your favorite and most trusted AI delays tracker.
according to projections, GPT-6.2 Astra was meant to launch last Thursday, and Fable 5.5 was meant to launch over two weeks ago. welcome to the doldrums
Show more
Mazebench vs Kimi K3.1 vs DeepSeek v4.1
Kimi K3.1 Pro is the first open weights model to have a nontrivial score, achieving 2%
DeepSeek v4.1 Flash is not far behind at 1%
@adcock_brett hark can fill out captchas and get you free pizza, codex won’t do either, thanks for pizza
MazeBench vs Gemini Argon 4
There’s a new kid on the block
Gemini Argon 4 scores 12%
It’s a 2nd place score, just below GPT-6 Astra
MazeBench vs GPT-6 Astra
Astra spent 60+ hours in this 3D open world spatial reasoning eval.
Final score: 14%
new MazeBench results are in
ox-alpha scores 0%
grok 4.6 scores 0%
glm-5.3 scores 0%
qwen 3.8-max scores 0%
gemini 3.7 flash scores 1% and does better than kimi k3
Introducing MazeBench, a 3D open world environment that evaluates long term planning with visual spatial reasoning.
It spans hundreds of rooms and puzzles. Today’s best agents cannot progress beyond the initial levels.
Show more
