i am new here some all llm needs api there is a free api unlimited ?
At the moment you can use Google Gemini for free if you create an API key.
Other then that OpenRouter with Qwen-Models are pretty cheap to use.
3 Likes
HuggingFace is free and decent on token use. Just sign up for several so you can swap them if you run into caps. Also, you can try experimental models for free, or of course run a model locally, completely free. Lots of options to get started.
2 Likes
Open router also have some free but not sure about rate limits etc.
For powerful free api using Gemini 2 flash go to aistudio dot google dot com then click on the left get api key then in the center of the screen click it again and boom your good to go. near endless creating for now at least.
2 Likes
You can get ChatGPT 4o for free through Microsoft’s Azure AI (GitHub). We had a discussion on this here. While Bolt.diy doesn’t support it yet, I submitted a PR (which might need to change due to some structural changes pending) and we should have support and docs soon.
My APIs suddenly stop working, both OpenAI and Open Router. Do you know how to fix this?
Both are active and have credits.
Hopefully you have fixed this. most of the time its just a hiccup or in some cases to many calls per minute or even over taxed context window.
Maybe rate limit? Did you get this working again or is it still an issue?
It stopped working, until yesterday I had to test several APIs from openai and open router, but today nothing works.
Did you update or change anything? Run out of credits (I would hope not)?
llama 3 8b is working but can’t move the project.
Yes, I have credits on openai and open router
I didn’t update, I followed the tutorial, they were failing from the beginning.
Ive experienced connection issues before, but only when using an IDE where I was trying to throttle more calls per minute than allowed. OpenAI has been really hit-or-miss for me as well, which is why I switched to Google Gemini.
While it may not be as powerful as OpenAI’s GPT-4 (01 preview), Gemini offers other advantages that ChatGPT lacks. For example, its context window spans 1 to 2 million tokens. This means I can input scripts as large as 500,000 tokens and still get a response, even if it takes 3 to 5 minutes.
Today, 12/23, the Llama Free APIs started working better and then stopped. And with great difficulty, they were able to make changes to ready projects and deliver the preview.
My experience has been very frustrating.
I’m thinking about creating a topic for super beginners, because without specific help, they will quickly give up.
1 Like
I do love a free API, so thought I’d share GLHF Chat . This platform supports most of the models available on Hugging Face, preloaded with around 20 of the latest and most popular open-source models. You can easily load additional models by simply entering their Hugging Face URL in the provided model loading box.
While it may not support some of the older models, that’s not typically a significant drawback. Best of all, GLHF Chat offers OpenAI-compatible API access for free, although they do plan to introduce a charge at some point in the future.
Hi @smokepigdad,
thanks for sharing. Someone else also mentioned this in another topic. I just can say the same here:
For me this website/provider looks not very trustworthy. No imprint, details, etc. So everyone has to decide themself if they want to sign up there.
For me this is no option at the moment, but happy to hear some other comments on this.
Hey there!
The website is gradually improving, and I’ve come across a few articles on tech websites discussing it. I believe they’ve been running it for the past six months not as a traditional business but rather to refine their setup, which might explain why it doesn’t yet have a professional appearance.
Personally, I haven’t been overly concerned about its trustworthiness since you only need to provide an email address to sign up and I don’t send anything I wouldn’t want others to see. However, I understand why some people might have reservations, especially if they’re sending sensitive information and seeking a guarantee of privacy. If they want to check out my attempts to transform obscure topics into Dungeons & Dragons classes, or to create new Star Trek centered around Janeway been put on trial for war crimes, the fact that it’s free means they are more than welcome to! 
SPD