Skip to content

Comment on GPT-4-turbo preliminary benchmark results on code-editing

Comments

Has anyone been able to access the 128k context window? I'm not seeing that option in the API playground

Yes, it’s fantastic. It is called gpt-4-1106-preview

You might want to double check. The latest turbo models for 4 and 3.5 don’t differentiate. There is only one model and it has 128k context.

I wonder if 3.5 / 4 is really just "agents are on" or not.

The reason he's not seeing it is because it's only available as an API endpoint. It doesn't show up in playground.

Nope, they definitely show up in the playground.

I think I need to be in the higher usage tier to unlock the full 128k context length

Confidently incorrect

Yes, it should be under gpt-4-1106-preview.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.