Add gpt-5-pro model cards for openai and openrouter providers

[AurelienRibon]

This powerful (but slow and expensive) model brings sota reasoning capabilities, it is crucial in math-heavy fields.

OpenAI model reference
https://platform.openai.com/docs/models/gpt-5-pro

OpenRouter model reference
https://openrouter.ai/openai/gpt-5-pro

[rekram1-node]

thanks!

[AurelienRibon]

@rekram1-node Thanks!
Now that we can use it, an interesting issue arises: the model only supports a high level of reasoning, and I support OpenCode defaults to medium.

opencode --model openai/gpt-5-pro run Hello
Error: Unsupported value: 'medium' is not supported with the 'gpt-5-pro' model. Supported values are: 'high'.

Would it make sense to introduce a new property to the model schema to define the supported reasoning efforts?
I can try to propose a PR for both OpenCode and models.dev if needed.

Of course, one can workaround this issue by defining a custom .opencode/agent definition, but being able to use the model without creating a custom file would be a much nicer user experience. At least allowing opencode to propose --reasoning-effort argument would solve the problem.

[rekram1-node]

opencode passes a default to got-5 models we need to exclude pro from that

[AurelienRibon]

Indeed I found the place, fixed here: sst/opencode#3474
And this one to let people customize the reasoning effort from the CLI, for increased flexibility: sst/opencode#3475

[rekram1-node]

Nice thank u!