gemini models support

[Krank41]

Does this repo support gemini models ??

[essiebx]

i believe only Claude-centric integration

[adamlu123]

Thanks for noting that! Feel free to make a pr @Krank41 if you would like, otherwise we can gradually rollout support for more models as well.

[Krank41]

Hi @adamlu123 i tired to code and use gemini model 3.5 flash model
i asked "Search for flights from SEA to JFK on 2026-08-15 to 2026-08-20" start-url https://www.google.com/flights
the agent is going with 50 steps and 15 screenshots so far and its not completed.
i think first time agent take more time to solve the tasks comparing to browser use(my current fav) .
i still think may be my Gemini implementation has bug i will look more closely and i will try to validate the accuracy you guys reported on Online-Mind2Web and Odysseys and also the time taken by agent for solving them . i do see in odysseys leaderboard top model is Gemini 3 Pro Preview so gemini 3.5 flash should work.. i will made the PR with gemini model once i validate the results you guys are getting on benchmarks.

[adamlu123]

Hi @Krank41, thanks for trying. Which yaml config are you using? local_browser.yaml typically takes less steps (11 steps with gpt5.4 with this task in my run). base.yaml usually takes longer step since the goal is to produce a reusable python script that can be reused.

[Krank41]

Hi @adamlu123 i have added the PR for gemini model.
please let me know the changes or problems in the code i am happy to change it and contribute to this project.
#38