🚨Elon Musk just said that Grok 4.5 (1.5T V9) is now in private beta at SpaceX & Tesla:
• Grok 4.5 is built on the 1.5T V9 foundation model
• It's apparently already close to, or possibly ahead of, Opus.
• RL training is still running.
Elon also said xAI plans to release completely new models trained from scratch every month for the rest of the year.
This is going to be incredible, will you be switching to grok and trying this model out, I also can't believe they are catching up this fast?
GPT-5.6 Sol is our most capable model yet for cybersecurity.
It shifts the performance-efficiency frontier for long-horizon security tasks including vulnerability research and exploitation.
GPT-5.6 Leaks: Coming on Thursday
- kindle-alpha is expected to be the launch checkpoint for GPT-5.6 / GPT-5.6 Pro, while gpt-bidi-1 may arrive as the new voice model.
- Vision and frontend generation are better with an image reference, it can almost recreate designs 1:1.
- SVG generation is one of its biggest strengths, especially 3D/static SVGs, where it can even outperform Fable 5.
- Its also very strong at game creation, producing cleaner visuals and more consistent results
- with the right prompts, GPT-5.6 Pro can even beat Fable 5, though Fable 5 still overall better.
- Playwright support is coming to ChatGPT for browser automation, and the knowledge cutoff is to be December 2025.
Gemini 3.5 Pro: Google Might Finally Be Cooking
- Logan Kilpatrick said the team is “cooking on 3.5 Pro!”, and Google taking extra time could be a good sign.
- Expect stronger vision, better multimodal reasoning, improved memory, more capable agentic workflows, and better SVG generation.
- A Gemini Super App experience and a new native image model (instant-ramen) could also arrive around the same timeframe.
- It will likely ship with stricter safety filters and content moderation.
- The biggest hope is that Google finally fixes the laziness on long and complex tasks seen in earlier checkpoints.
- My current expectation is a release around June 30.
Seed 2.1 Pro Preview ranks #8 in Code Arena: Frontend, scoring 1539, on par with Opus 4.6. It performs strongly on React apps and lands in the top 10 for five of seven subcategories. In those areas, only a handful of frontier labs rank above it: @Zai_Org's GLM-5.2 and @AnthropicAI's Claude models.
Highlights:
- #7 on the React leaderboard, #14 on HTML
- #6 Brand & Marketing
- #9 Content Creation Tools, Data & Analytics
=#10 Reference Based Design, Consumer Product
This is an early access preview of the model. It will be publicly available in a few weeks.