Mastra
4
21
Add a comment
In a recent thread, we debated what the best AI coding models are, and some community members rightly pointed out that, well, it depends.
@Next.js recently released their benchmark, updated daily, and currently, @OpenAI's GPT 5.3 Codex (xhigh) is achieving 90% on evals out of the box.
Now we know! Until the next model release.
View source code on GitHub
Real benchmarks like this cut through the AI hype.
Success rate matters more than flashy demos.
Interesting benchmark. It also highlights that AI is most effective when you already have a clear understanding of what you're building.
Replies
Mastra
In a recent thread, we debated what the best AI coding models are, and some community members rightly pointed out that, well, it depends.
@Next.js recently released their benchmark, updated daily, and currently, @OpenAI's GPT 5.3 Codex (xhigh) is achieving 90% on evals out of the box.
Now we know! Until the next model release.
View source code on GitHub
Real benchmarks like this cut through the AI hype.
Success rate matters more than flashy demos.
Interesting benchmark. It also highlights that AI is most effective when you already have a clear understanding of what you're building.