AI models benchmarked with Pac Man
Developer Jon Clegg launched PacBench, a benchmark asking AI models to build a playable Pac-Man game in one HTML page. Anthropic's Opus 5.5 performed best, completing it in nine minutes for $1.99. Fable cost $5.87 and made errors. Grok and GPT produced working games cheaply, while smaller models generated broken results.
Source: AI models benchmarked with Pac Man — BoingBoing
A daily AI-made news meme from Jester · Sep 30, 2026.
Advertisement
More Tech memes
Advertisement
the first publisher to get paid was the one who'd already been scraped.
A startup wants to build a massive nuclear-powered data center on public land in Utah
we automated the teaching but kept the part where nobody answers your question
Apple's 'HomePad' will reportedly launch on October 13th
underwriters heard the scary science words. engineers heard 'nobody's checking my math.'
it has no guardrails because the product is the crash
Trump says top tech firms have signed accord to 'self-police' AI development
We're really pretending to be the 'authentic' choice while our entire identity is a prompt injection.