Add more models (4.1 and 4o)

[samwega]

This plugin is amazing, it can replace the functionality of type.ai which is what I've been waiting for in obsidian for months. However, it is a little short sighted, this can be used to get text improvements that go far beyond simple proofreading of grammar and syntax.

Of course, you would need to change the prompt, which I have, but you also need access to better writing models, particularly 4o and 4.1. Can we please get these in the dropdown menu?

[chrisgrieser]

Thank you.

Yeah, the plugin really is only intended to be used for something like proofreading, since suggestions for larger rewrites would probably require a very different UI. As such, I have only included 4.1 mini and nano, since they are best suited for proofreading, namely through speed & low price, with reasoning not being needed for tasks that only requires grammatically and orthographically correct output.

I can see that some people nonetheless might want to tweak the prompt to do things beyond proofreading, so I added gpt 4.1 to the choice as well in release 0.15.4 (should be available in the community store any moment). Bear in mind though that gpt 4.1 is literally 20 times more expensive as 4.1 nano, so I'd only use it if really needed.

I did not include the reasoning models, since I do not think they are needed for tasks like rewriting text. Inspecting the specs of the 4o models, they are all strictly worse than their 4.1 equivalents, so I don't think there is a point in adding them either. I am trying to keep the plugin as simple as possible for non-technical users—other plugins often just throw all available models at you, even if it is a whisper model and the plugin only does text writing, which I think is just poor software design.

If you have a specific model in mind and can explain why it would make sense to add it, I'm totally open to adding is as well, though.

[samwega]

That was super fast, thank you!

I had already edited my local plugin to add those two models for testing (I used Gemini, I can't code). The reason I recommended 4o too is that while it's not as good as 4.1 for general stuff, it's appreciated by writers. Professional writing apps, like Type.ai only use 4o and Claude, as the two best creative writing. That being said, it is entirely possible that 4.1 is dethroning 4o even for creative writing, I think the jury is still out (?). I guess 4.1 is enough, unless more people ask for 4o.

Since you seem to be particularly concerned with pricing, I would recommend, should you ever want to expand beyond OpenAI, Gemini 2.5 Flash - the best bang for your buck probably at this time. I also think writers will eventually ask for Claude, even tho I don't particularly like it myself.

One problem, when you start introducing more models, especially from more companies, is that different models respond to different prompts. So a future feature that I think would make sense is some kind of multi-prompt selector, for custom prompts. I would suggest adding a submenu in the right click menu for selecting among prompts. Too complicated?

P.S. I am currently using this prompt with Proofreader (still being fully tested, so far so good):
Act as an expert proofreader and writing assistant. Review the following text for minimal yet impactful improvements. Enhance clarity, conciseness, readability, grammar, and language. Crucially, preserve the original meaning, personal style, and any technical jargon. Suggest structural changes only if they significantly improve flow or understanding—avoid unnecessary expansion or major reformatting (e.g., no unwarranted lists). Focus on high-value corrections and refinements that maintain a natural, cohesive feel. If the text is already effective, make minimal or no changes. Do not alter the core information. Output only the revised text, with nothing else. The text is:

P.P.S. I was paying $29 a month for type.ai until today, your plugin made me cancel the subscription, because the only feature I wanted from there was this inline granular diffing which no other app offered, until Proofreader.

[chrisgrieser]

Professional writing apps, like Type.ai only use 4o and Claude, as the two best creative writing. That being said, it is entirely possible that 4.1 is dethroning 4o even for creative writing, I think the jury is still out (?).

gpt 4.1 is only 1-2 weeks old. I assume the other writing apps will add support for it soon. By all specs, it seems to be superior to the 4o models, so yeah, I don't think you should use 4o anymore.

Since you seem to be particularly concerned with pricing,

I personally am not particularly concerned, since even with the expensive models, you need to do a lot of proofreading to pay much. However, I simply find it irrational to use a more expensive model when the simpler model does the job just as good. More importantly, the flagship or reasoning models are also slower than the mini/nano models, which is particularly noticeable when you proofread a whole document at once.

I would recommend, should you ever want to expand beyond OpenAI, Gemini 2.5 Flash - the best bang for your buck probably at this time. One problem, when you start introducing more models, especially from more companies, is that different models respond to different prompts. So a future feature that I think would make sense is some kind of multi-prompt selector, for custom prompts. I would suggest adding a submenu in the right click menu for selecting among prompts. Too complicated?

Not complicated, just a lot of busywork. Every provider expects a different request, and given that there are dozens of providers, I am not keen spending hours to add support for all of them. On top of all that, every additional provider also means extra maintenance work (update model lists when they publish new models etc.). Given that I am creating and maintaining this plugin in my free time, on top of my full-time job and various other projects, my time is simply a limited resource I want to spend somewhat wisely.

P.S. I am currently using this prompt with Proofreader (still being fully tested, so far so good):
Act as an expert proofreader and writing assistant. Review the following text for minimal yet impactful improvements. Enhance clarity, conciseness, readability, grammar, and language. Crucially, preserve the original meaning, personal style, and any technical jargon. Suggest structural changes only if they significantly improve flow or understanding—avoid unnecessary expansion or major reformatting (e.g., no unwarranted lists). Focus on high-value corrections and refinements that maintain a natural, cohesive feel. If the text is already effective, make minimal or no changes. Do not alter the core information. Output only the revised text, with nothing else. The text is:

I like this prompt, I might yoink one or sentences for the default prompt if you don't mind.

P.P.S. I was paying $29 a month for type.ai until today, your plugin made me cancel the subscription, because the only feature I wanted from there was this inline granular diffing which no other app offered, until Proofreader.

Haha, thanks, do you mind if I quoted with this in the readme as "testimonial"? 😆

[samwega]

Sure bro, use all that freely!

I agree with most of that. Regarding 4o, so far people are saying it is better than 4.1 when it comes to creative writing / brainstorming, but I have to use it more myself to see what I prefer. Anyway, I've added 4o for myself and will be testing.

The thing is, nobody is investing in writing models, all the new models are focused on coding and other kinds of work. 4.1 is so much better at following instructions, but it's more robotic when it comes to being creative. It's the same with Claude imo, 3.5 is better than 3.7.. there's not much money in developing better writing models right now so 3.5 and 4o are peak for now, for that one task. If you want to write academic stuff or just notes, certainly 4.1 is going to win.

As for the thinking models, I agree, they wouldn't work for your plugin and make no sense. Plenty of other plugins use thinking models, for generating content, but it doesn't work for proofreading or correcting text, too slow and thinking models are not better at writing text, it's not what they're made for, they're only better at coding.

Anyway, I'm pretty satisfied with what your plugin does right now, I use others for other stuff, no point trying to do everything. And maybe other people will step in and work in Gemini and Anthropic in time. I use Gemini via Gemini Scribe, it's a great plugin, but completely different purpose to this. Thank you again!