I will explain the strengths of each AI out there that I use, so you can know exactly which AI models/tools are the best in my opinion! This is what these AIs are good at, in my opinion! Text Generation (LLMs): -GPT-4.1 by OpenAI (ChatGPT): •Questions in general •Web search

The ex-director of AI at Tesla and one of the members of the founding team of OpenAI, Andrej Karpathy, just released Nanochat. Nanochat is a full-stack training/inference pipeline to train your own AI (LLM). By training it for a few hours, you would be able to talk to your own

If you are a developer who uses Perplexity’s Search API, you can now filter searches by specific domains. For example, if you only want Wikipedia articles, you can use it.

OpenAI just partnered with Broadcom to develop custom AI accelerators. OpenAI will design the chips and Broadcom will build and deploy them. OpenAI is collaborating with a lot of compute providers in the past few weeks

Moondream 3 Preview is now available on FAL AI. It's an AI vision model for captioning, etc.

Scira AI now shows you the sources in a list format, so it's much easier to navigate through the used sources.

I just got access to the best AI video generation model (Sora 2 from OpenAI). If anyone wants to try it for free, here is an invite code: 4NS6S7 It's really good.

Google released Gemini 2.5 Computer Use. It's a model that can take actions and navigate in a browser.

xAI released Imagine v0.9, their new video generation model. It's much better than Imagine v0.1 and is completely free right now. You can use it in the Grok app.

Cursor (AI code editor) now has a plan mode. It will gather information and write a detailed plan before changing your code.

Sora 2 and Sora 2 Pro (OpenAI's AI video generation models) are now available in the API, so developers can integrate video generation into their applications. OpenAI also released a cheaper image and audio generation model, for developers.

If you are a developer who has integrated GPT-5 into your application, you can now get 40% faster responses with the priority tier.

OpenAI released AgentKit for developers. With Agent Builder developers can create AI agents without writing code and embed the agent into their application with ChatKit, while having Guardrails integrated. Agents that took developers days of development can now be made in a few

ChatGPT can now interact with apps like Spotify and so on, so it can take actions.

OpenAI will probably release a new feature for developers soon called Guardrails. It will give developers that have integrated GPT models in their apps the ability to add moderation, hallucination detection, etc.

Perplexity just launched their Search API, which means that developers can integrate Perplexity's search engine into their apps. It takes 0.4 seconds to return results while beating Google when it comes to quality.

FAL AI just added Wan 2.5 to their model list. Wan 2.5 is one of the best AI video generation models right now. The cost is 0.5 USD per video.

Suno AI just released their newest AI music generation model Suno V5. It's better at generating voices that sound human and it gives you more control over the music it generates.

Perplexity AI just introduced Email assistant. It's an AI agent that can draft replies, schedule meetings and much more. It's currently only available to Max users, so users who are paying 200 USD per month.

OpenAI's new coding model GPT-5-Codex is now available in Cursor (AI code editor). It thinks only for a few seconds for simple problems, while it thinks longer for harder problems, so it changes the reasoning (thinking) effort based on the task type.