I just got access to Claude 3 which is finally available in Canada it appears as of today. I've been using Claude 3 via POE and perplexity to bypass the Canadian restrictions. It's amazing to see it available in Canada.
It has 3 models, Haiku (meant to be a chatbot or simpler tasks), Sonnet (middle model) and Opus (top model, highest resources and slower)
Sonnet is currently free to use and performs nearly as well as GPT 4 from my tests and in some ways even better. I highly recommend it. Opus is actually better than GPT4 on benchmarks, and does coding even more efficiently without as many errors. Claude 3 is also a lot more relaxed than its predecessor which was notoriously and hilariously censored. The API offers 200K token context as well. Sonnet /Opus also seems more fresh and provides more detailed answers (GPT4 has a lot of nuances and issues with laziness that you can easily spot when using it extensively daily). I will be testing it extensively now that I have direct access.
Grok also released it's new model yesterday, available to Twitter premium subscribers, it appears to perform quite well, nearly on par with GPT 4. And it appears to be uncensored as well (see videos on Matthew Berman's tests below)
I thought this would also be a good time to give some additional AI related information and resources for anyone looking for more tutorials or want to get deeper into the subject, want to be in the know for newest updates.
The best channel for AI news
https://www.youtube.com/@mreflow
2nd one (don't let the obvious bait titles fool you, it's a great channel).
https://www.youtube.com/@WesRoth
If looking for tutorials on AI image generation these are your best options
https://www.youtube.com/@sebastiankamph
https://www.youtube.com/@OlivioSarikas
Local LLM fine tuning, tutorials (if you want to know how to create your own uncensored AI GF chatbot or roleplay bot, this is your best source to know this)
https://www.youtube.com/@Aitrepreneur
These 2 are more advanced by programmers who dive deeper and do extensive testing of all LLMs, very valuable resources if looking to build AI aided programs, agents, etc.
https://www.youtube.com/@matthew_berman
https://www.youtube.com/@AllAboutAI
For music AI, by far the best is Suno, I have access to V3 and the music generated is absolutely mind blowing
https://www.suno.ai/
For voice generation, use ElevenLabs
https://elevenlabs.io/
For image generation, your big 3 are:
Midjourney (closed source, censored, subscription, amazing photorealism and best quality images overall)
Dalle 3 (neutered photorealism, most coherent model, can be used for free via Bing / Copilot or as part of a chatgpt subscription)
Stable Diffusion (models were released to public and can be run locally, and likely the source of all NSFW AI imaged on the internet right now, SD3 is about to be released soon which is close to Dalle 3 in text generation)
For video generation, until Sora is released, the best text to video and image to video tools are
Runway
https://runwayml.com/
Pika Labs
https://pika.art/
Both are good, but I wouldn't call commercially viable at the moment.
Hopefully some of this helps for anyone who's curious.