Astra is insanely powerful. That’s exactly why we should be precise about what it is.
I went through the AGI claims, the benchmarks, and a real failure in my own codebase.
My conclusion: the rhetoric is ahead of the evidence.
Anthropic says Claude helped generate machine-checked proofs for 239 fuzz targets across 123 open-source libraries.
Under specific assumptions, those programs were proven not to crash.
That's considerably more interesting to me than another benchmark showing how fast a model writes code.
33
12.8% on SWE-sweep with the mini-SWE harness.
39% when running in Antigravity.
Same model, very different results depending on the setup.
The harness might matter just as much as the model itself.
Gemini 4 Neon is rolling out now.
Sitting between Carbon and Hydrogen, it brings frontier intelligence and fast task completion for complex software engineering, knowledge work, and science.
1
4
57
Well, GPT-6.2 Astra finally dropped!
OpenAI is claiming better performance at less than a third of the cost per task compared to GPT-6 Astra.
Now I want to see how it holds up against Fable 5.2 and Opus 5.5 on real projects.
Time to put it to work.
Replying to @OpenAI
GPT-6.2 Astra—a frontier model built for your hardest projects. It sets new highs on evals across professional work and scientific research at less than a third of the cost per task of GPT‑6 Astra.
You’ll see the difference in things like slideshows, websites, and 3D designs.
1
2
59










