🦙 Meta Drops Llama 4

The new open-weight model family is out, and it’s impressive.

Key Details:

• Trained on over 15T tokens with a 128k context window

• Scores better than Gemini 1.5 Pro and Claude 3 Opus on many tasks

• Comes in two sizes: Llama 4 8B and 70B

• Licensed for commercial use with attribution

• Meta is building a 400B-parameter model, expected later this year

Why It Matters

This is Meta’s biggest open-weight release yet. It shows that open models can compete at the frontier, especially for devs and startups who want control. With Llama 4, the open ecosystem just got more credible and more capable.

🧠 Copilot Just Got More Useful

Microsoft’s assistant can now remember you, help with bookings, and complete purchases.

Key Details:

• Copilot now saves personal context like preferences and habits

• Can complete tasks like shopping and scheduling with minimal follow-up

• Works across Microsoft products and partner platforms

• Microsoft says this moves Copilot closer to a true assistant

Why It Matters

This is where agents start to feel useful. Copilot is no longer just a tool that responds to prompts. It’s a system that remembers, adapts, and acts. For creators, that means more reliable assistance across workflows.

🎨 Midjourney V7 Is Live

The image model is faster, cleaner, and supports voice prompts.

Key Details:

• Voice prompt support is now available in the Discord server

• Faster image generation and improved coherence

• All-new model architecture under the hood

• Midjourney says V7 will be followed by inpainting and video tools later this year

Why It Matters

Midjourney V7 is the most polished yet, but the real story is momentum. It’s evolving from a vibe generator to a serious creative engine. Voice input and speed improvements make it more usable, especially for creators on the move.

🔜 OpenAI Delays GPT-5

But lighter, more nimble models are dropping soon.

Key Details:

• GPT-5 is now expected later this year

• Two new models, o3 and o4-mini, will arrive sooner as standalone tools

• These are meant to improve inference speed and cost

• Sam Altman says the delay is about getting it “really right”

Why It Matters

OpenAI is pivoting to modularity. Instead of waiting for one big launch, they’re releasing a stream of focused models. That keeps the platform fresh and competitive across use cases, especially for startups and API users.

🌐 Gemini 2.5 Pro Opens Up

Google’s best model is now live in Vertex AI and open for API use.

Key Details:

• Gemini 2.5 Pro is Google's most advanced model yet

• Multimodal, long context, and deeply integrated with Google services

• API now open to all developers, with pricing similar to GPT-4

• Chrome DevTools now include built-in Gemini code assistance

Why It Matters

Google has been quiet lately, but this puts them back in the race. Gemini 2.5 Pro is powerful, accessible, and ready to plug into real products. If you’re building for scale, this is a serious contender.

🕐 Quick Bits

📦 Perplexity launches “Pages”
A new feature for turning AI searches into sharable documents or blog-style posts.

💡 Claude 3 gets Code Interpreter
Anthropic quietly added tool use to Claude 3 Opus for pro users.

📽️ Kling’s open beta expands
The AI video platform is now accepting more users for testing.

🛠️ Trending AI Tools for Creators

• 🦙 Llama 4 – Meta’s best open model yet, with 128k context
  🎨 Midjourney V7 – Fast, expressive, and now with voice input
  🧠 Copilot Recall – Personal memory for smarter AI help
  🔍 Gemini 2.5 Pro API – Google’s top model, now open to devs