Last week the Anthropic team unveiled their most advanced AI model yet: Claude 3.5 Sonnet. Our team hustled to add 3.5 Sonnet to Sudowrite, and it's now available as an Experimental model.

In benchmarks, Claude 3.5 Sonnet is proving more capable than Claude 3 Opus and even OpenAI's new GPT-4o model. But what does that mean for your prose? Here are some early reviews of the model from members of the Sudowrite Discord community:

"I LOVE Sonnet 3.5, personally, and it's awfully cheap for what I get out of it. ...Sonnet 3.5 will actually $#!%ing finish sequences and when I tell it that I want things to happen 'over the next several days' it will actually pay attention to it." - Brett

"I adore the new Sonnet 3.5 and havent had any trouble with it repeating." - Stacy

"Sonnet 3.5's ability to understand logic is remarkable - compared to every other model, it nearly NEVER gets logical details of my scene wrong. It's consistent with past details in the scene, and with the characters personality and past emotional reactions to similar situations. I'm shocked by that aspect of it, personally. It just 'gets' what is happening without needing explanation or lots of direction (vs other models)." - SnowBloom

"I am loving Sonnet 3.5." - Christina

You can try Claude 3.5 Sonnet right now by selecting it in your Write Settings, Chapter Generator, or Advanced Plugins.

One other note regarding available models:

Our team has confirmed that support for Airoboros has been discontinued, and as such we've had to remove it from the Write button and Chapter Generator's Experimental models section. If you previously selected Airoboros to power a Plugin, you'll need to edit that Plugin and select a different model to keep it functional.

We'll do an audit in the near future to delist any Plugins that are still using Airoboros, to decrease the chances of folks finding a broken Plugin—but in the meantime, you can use the 🚩 Report function to flag those to our attention.