Update gemini models config

[reynaldichernando]

for some reason the numbers we have are inaccurate, compared to google own docs
i found these while adding the gemini 3 flash

• gemini 3 pro preview should have 65,536 max tokens (source: https://ai.google.dev/gemini-api/docs/models#gemini-3-pro)
• gemini 2.5 pro should also have 65,536 max tokens (https://ai.google.dev/gemini-api/docs/pricing#gemini-2.5-pro)
• gemini 2 flash should have 1M context (https://ai.google.dev/gemini-api/docs/models#gemini-2.0-flash)

please help cross check and review this

although nothing is wrong/broken yet with these models?
they still work, but idk what the implication is having these incorrect information

[reynaldichernando]

hi @ProgrammerIn-wonderland , could you help to review this PR, thank you 🙏

[ProgrammerIn-wonderland]

This looks fine but I want to make sure there aren't weird effects with max_tokens first since max_tokens means something completely different for Gemini (maximum output tokens) than it does for the OpenAI models (max tokens in conversation). I think this change has the potential to cause issues

[reynaldichernando]

noted, btw for the gemini 3 flash, i followed the number in the docs, which shows 65,536

I want to make sure there aren't weird effects with max_tokens

okay, i'm not sure how the internals work yet for this, but just bringing this to your attention 🫡

[Salazareo]

so I think these should be good to be different than openai ones, but i won't merge in case @ProgrammerIn-wonderland finds something off with it

[ProgrammerIn-wonderland]

Yeah the issue here is still that we associate Max tokens with context length internally and do math based on that, we probably need to do a secondary, alternate calculation to properly follow geminis.