Instruction following. Opus 4.7 is substantially better at following instructions. Interestingly, this means that prompts written for earlier models can sometimes now produce unexpected results: where previous models interpreted instructions loosely or skipped parts entirely, Opus 4.7 takes the instructions literally. Users should re-tune their prompts and harnesses accordingly.
We appreciate your request to bring Claude Opus 4.7 to Antigravity. We will be submitting an internal feature request and will provide updates as they become available on the console.
I’d hope Google is prioritizing a usable product release before expanding “functionality.” The promise of frontend QA capabilities drew me to Antigravity for a trial run. What an absolute disaster. Retrying prompts endlessly due to API timeouts and when I manage to make any progress, the frontend tool is “unavailable.” I’ll stick with Claude- no need to wait for a new release.
Can you also please submit something to fix the server issues everyone is talking about and dealing with? It’s kind of annoying and wastes our credits. I’ve noticed since these server errors I run out of my plan A LOT faster and then have to wait hours to work more…It’s crazy to pay $250/month to be constantly kicked off of the connection due to the server apparently being overloaded.
yeah, it … paying $250/mth and being put on the shelf… despite it being on vertex. Heck even other IDEs have it running already. What happened to Gemini 3.1? Do they even care anymore?