model functionnal ?

[fredmo]

Hello,

model="gpt-oss-120b" is no more working.

Could it be possible to have model="best" or model="default" , or "precision" or "quick", or other thing else ?
( and without cloudfare )

The idea is that regularly the model used is no more functional, and instead of looking in the https://github.com/xtekky/gpt4free/blob/main/g4f/models.py for new model and test them until finding a functional one, it would be very nice to have an advice or a category of model validated and functional by default ?

Or an automatic script to test all the models available and share a benchmark result ? OK / KO

Thank you,
Frederic

[github-actions]

Bumping this issue because it has been open for 7 days with no activity. Closing automatically in 7 days unless it becomes active again.

[hlohaus]

Hello @fredmo

Thank you for your feedback on GPT4Free—it's helpful to hear about issues like the "gpt-oss-120b" model no longer working reliably.

Currently, GPT4Free provides two built-in default models:

• g4f.models.default (for standard text-based interactions)
• g4f.models.default_vision (for vision-enabled tasks, like image analysis)

You can import these directly from g4f.models in your code, or simply omit the model parameter (or pass an empty string "") when creating a request. When using a default model, GPT4Free automatically selects from a predefined list of available providers, rotating through them to find a working one. This helps avoid issues with individual providers failing (e.g., due to Cloudflare blocks or downtime) without needing to manually test models from the models.py file.

Regarding your suggestions for options like model="best", model="default", model="precision", or model="quick"—these aren't implemented yet, but they're great ideas! A "best" or "precision" mode could prioritize higher-quality providers, while "quick" could favor faster (but potentially less accurate) ones. We could also explore Cloudflare-free defaults or an auto-testing script to benchmark and report OK/FAIL status for models/providers periodically (e.g., via a GitHub workflow that shares results in the repo).

Feel free to share more details on your use case or ideas for new default models—we're always open to contributions or pull requests to improve reliability!

Best regards,
The GPT4Free Team

[github-actions]

Bumping this issue because it has been open for 7 days with no activity. Closing automatically in 7 days unless it becomes active again.