New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Which AI are you using recently?
Although ChatGPT started the AI wave, to be honest, I have been using Google Gemini and DeepSeek for the past year. Gemini API supports free individual calls, and DeepSeek can see the reasoning process. Both are smart enough to meet my daily work needs. What do you think?
Comments
Personally, I've been trying to use self-hosted models recently as they got somewhat capable now.
I have phi4, gemma3 and qwen3 on ollama+openwebui and honestly they're enough for most tasks.
If I need something smarter I usually ask free version of Claude 3.7.
deepseek is good because it's really easy to jailbreak
Yes, bro, Claude is also a very good AI. I also tried to deploy Ollama locally, but because my graphics card only has 8GB of video memory, it can only fit a model of about 7B, which is obviously not enough to meet my needs, but it is a very interesting process as an attempt.
Indeed, I really enjoy watching its reasoning process, just like watching a smart person solving a problem.
Been trying out Grok.
Good choice, I'm planning to learn more about it
grok has been working really well for me recently. i prefer it over my chatgpt and claude subscription.
Thanks, I'll give it a try.
Again the 'X' factor from Elon Musk. Has it been a quiet satisfying experience? If so, I would definitely give it a try.
Cursor AI is offering free pro for a year for students.
First I use cGPT, than I use Claude Sonet to correct cGPT mistakes and than I use xAI to finalize and optimize.
Gemini 2.5 Pro via AI Studio, quite good for programming, haven't encountered query limits compared to the main Gemini page.
Yes, bro, and you can use the API to call the development version model, such as the latest gemini-2.5-pro-preview-05-06.
Ollama with llama3.4 + qwen2.5-code
What model sizes (in billion parameters) do you run?
8B qwen3, 12B gemma3, 14B phi4
My GPU has 12GB vram
Claude for code (don't call me a vibe coder), ChatGPT for everything else.
back and forth with grok and gemini, lately trying new qwen 3 it seems promising
deepseek-reasoner/deepseek-chat, both local and official api. in my experience, gemini keeps yapping too much when i prompt to refactor my already fairly complex rust code. claude works great but too expensive when deepseek exists, and i’d rather avoid us companies in this regard. mistral works pretty well as 4o-mini/text-embedding-3 alternative.
Does anyone know which models have the most generous free API usage at the moment?
I have a fairly big text classification job, (categorising ~8K recipes for mealie), but I didn't pay for the recipes and don't want to pay to classify them either 😜
Gemma 3 is the best for this (14k request per day), but it lacks structured output support. You can do that by a prompt or use Gemini 2.0 Flash which is limited to 1500 request per day.
xAI also offers $150 per month free credits
Subscribed to ChatGPT, use it for general stuff and general image generation.
Also use Google Labs Image FX for free image generation, I find it gives more specifics on what you ask for.
Use Gemini and DeepSeek interchangeably when I want a different take on something.
Use Claude Sonet for content/article generation.
Have been using Grok quite a bit lately as it’s very fast and especially good with tech/coding stuff.
Chatgpt free for most questions, then use Gemini to double check.
Try Gemini (can’t remember what the base model is called) preview models. Free api use but some rate limits.
This looks perfect. Thanks for that!
Model: Gemma 3
Request Per Minute: 30
Tokens Per Minute: 15,000
Requests Per Day: 14,400
It'll process my 8K recipes in a few hours, and I'll have enough requests to do them individually so structured output won't matter.
t3.chat for every model for $8
gemma 3 (unsloth) on my macbook for flights
chatgippity
GigaChat bot in Telegram for:
1) Rewrite text to check grammar and style.
2) Small addition/modification in code.
3) Generation images for banners etc.
It is work well for Russian language.
Grok 3 at most, it's smart enough
Been using Gemini 2.5 Pro more and more lately. Others are o4-mini and Grok Mini, and sometimes Claude 3.7.
All via APIs to my selfhosted Open Web UI.