Ancestors

Toot

Written by TTimo on 2025-01-04 at 19:49

Each of these LLMs has a (roughly) $20/month subscription plan - which one do you use or recommend? Best value and future prospects for the money in your opinion?

[#]ai #llm

=> More informations about this toot | More toots from TTimo@mastodon.social

Descendants

Written by mikołajczyk, podobno on 2025-01-04 at 20:01

@TTimo This is a ragebait for fediverse users, right?

=> More informations about this toot | More toots from mkljczk@fediverse.pl

Written by ROTOPE~1 :yell: on 2025-01-04 at 20:08

@TTimo I'll go for either Bixby Bintelligence, or Huawei Huawisdom.

=> More informations about this toot | More toots from rotopenguin@mastodon.social

Written by Hojjat :ferris: on 2025-01-04 at 22:46

@TTimo As far as I know Copilot also has Claude Sonnet as one of the models you can choose with it. (not sure if you need to pay more or not)

=> More informations about this toot | More toots from hojjat@fosstodon.org

Written by TTimo on 2025-01-04 at 22:56

@hojjat yeah I guess Copilot is mostly Claude in different clothing. The Copilot web UI seems worse to me though, but I like that it puts web links in it's answers, which the others don't seem to immediately do.

=> More informations about this toot | More toots from TTimo@mastodon.social

Written by TTimo on 2025-01-06 at 16:51

@hojjat actually I was wrong .. I think the Microsoft stuff uses OpenAI .. so ChatGPT, not Claude but same idea

=> More informations about this toot | More toots from TTimo@mastodon.social

Written by Hojjat :ferris: on 2025-01-06 at 23:13

@TTimo Yeah. But they did add claude sonnet recently too: https://www.anthropic.com/news/github-copilot

Unfortunately I haven't been able to get it to work in Nvim, but in vscode you can choose it from a drop down menu apparently.

=> More informations about this toot | More toots from hojjat@fosstodon.org

Written by Debby on 2025-01-05 at 16:28

@TTimo I use Jan ui locally by using mostly free models and occasionally switching to Copilot, my monthly expenses dropped.

=> More informations about this toot | More toots from debby@hear-me.social

Written by lainy on 2025-01-06 at 16:23

@TTimo kagi is a good deal too

=> View attached media

=> More informations about this toot | More toots from lain@lain.com

Written by TTimo on 2025-01-06 at 16:49

@lain oh that's https://help.kagi.com/kagi/ai/assistant.html .. never heard of it, that's interesting

=> More informations about this toot | More toots from TTimo@mastodon.social

Written by Thibault Mol. 💤 from FOSDEM on 2025-01-06 at 19:19

@TTimo @lain

(have been recommending Kagi even before I was a consultant for them, disclaimer).

But yeah, Kagi Ultimate makes a lot of sense. It's not perfect when compared to having some of these llm's directly. But considering you're getting the best search engine included, the price is hard to argue against.

=> More informations about this toot | More toots from thibaultmol@en.osm.town

Written by TTimo on 2025-01-08 at 14:53

@thibaultmol @lain fwiw I've been playing with kagi for a few days now and I notice the same problem: https://kagifeedback.org/d/2093-quick-answer-follow-up-questions-assistant-integration

=> More informations about this toot | More toots from TTimo@mastodon.social

Written by TTimo on 2025-01-08 at 14:54

@thibaultmol @lain plus you have to pay for the plan first to get access to the assistant and have some opportunity to see how it is able to continue the conversation.

=> More informations about this toot | More toots from TTimo@mastodon.social

Written by Thibault Mol. 💤 from FOSDEM on 2025-01-08 at 15:10

@TTimo @lain personally I just upgraded for one month to try it out and then stuck with it. But yeah, I feel you.

I should mention that in the New year's celebration live stream they did mention that the lower tier plans are also going to get some basic level of access to the assistant iirc

=> More informations about this toot | More toots from thibaultmol@en.osm.town

Written by Thibault Mol. 💤 from FOSDEM on 2025-01-08 at 15:12

@TTimo @lain and about this specific feedback item.

In general I never really use the 'continue in assistant' button. Currently a basically just opens the URL in the assistant and that's it... It doesn't actually work with the input you saw in quick answer.

So if I know that I'll probably prefer ai for a question, I'll just ask it in the assistant directly.

=> More informations about this toot | More toots from thibaultmol@en.osm.town

Written by Thibault Mol. 💤 from FOSDEM on 2025-01-08 at 15:12

@TTimo @lain Do note that as a non-assistant tier subscriber you do have access to fastgpt https://kagi.com/fastgpt it's not as good as the actual assistant, but it should come included with your plan, so there's that

=> More informations about this toot | More toots from thibaultmol@en.osm.town

Written by Jonathan Schofield on 2025-01-06 at 16:25

@TTimo I have voted by choosing none.

https://mastodon.social/@urlyman/113701656729629952

=> More informations about this toot | More toots from urlyman@mastodon.social

Written by Glyph on 2025-01-06 at 19:21

@TTimo really needs a “none of the above” :)

=> More informations about this toot | More toots from glyph@mastodon.social

Written by TTimo on 2025-01-06 at 21:18

@glyph yeah that's my bad. There's a sizeable anti AI crowd on the fedi of course, but I didn't anticipate them to be so virtue signaling about it.

I'll probably just use a content warning tag moving forward.

=> More informations about this toot | More toots from TTimo@mastodon.social

Written by Glyph on 2025-01-06 at 22:07

@TTimo I am certainly a part of the "anti-AI" crowd but just for the purposes of this poll — avoiding any normative "virtue signaling" statements — these services all strike me as quite bad value for money. To the extent that I would want to use a model, I would want something I can run locally.

=> More informations about this toot | More toots from glyph@mastodon.social

Written by Druid 🏴󠁧󠁢󠁷󠁬󠁳󠁿 on 2025-01-23 at 14:12

@TTimo I've only used ChatGPT and Gemini. I would have chosen ChatGPT. It's far superior to Gemini when it comes to coding. Gemini gives very poor solutions and gives up easily when trying to work with it to fix an issue. ChatGPT just keeps on trying until you get what you want. It feels more like a collaboration, like chatting to a friend in work. I tried Gemini again recently too see if it had improved but no.

=> More informations about this toot | More toots from druid@toot.wales

Written by TTimo on 2025-01-23 at 14:26

@druid yes, regarding code I had observed that ChatGPT seemed better overall. I don't use the LLMs as code assistants much atm, but mostly as web search helpers. Based on the discussions in this thread I went with @kagihq and it's been a very nice experience.

=> More informations about this toot | More toots from TTimo@mastodon.social

Written by Druid 🏴󠁧󠁢󠁷󠁬󠁳󠁿 on 2025-01-23 at 14:51

@TTimo @kagihq I hadn't heard of Kagi before. Good to know.

=> More informations about this toot | More toots from druid@toot.wales

Proxy Information
Original URL
gemini://mastogem.picasoft.net/thread/113771819608444869
Status Code
Success (20)
Meta
text/gemini
Capsule Response Time
423.501021 milliseconds
Gemini-to-HTML Time
5.536707 milliseconds

This content has been proxied by September (3851b).