I usually expect tech news to quiet down near the holidays, but I was blown away by this update. This AI professional just broke down a massive list of releases that proves the industry isn’t sleeping at all.
🧠 GPT-5.2 is Here
The expert highlighted the surprise release of GPT-5.2. It appears OpenAI focused heavily on math and science benchmarks to fix recent error rates found in previous versions. The creator noted a massive 400,000 token context window, allowing for roughly 300,000 words of combined input and output.
He also mentioned a major partnership with Disney. The original poster speculates this means we will soon see official Disney IP, like generating characters or scenes, directly inside tools like Sora.
🎥 Runway Gen 4.5 Testing
This video creator put Runway’s new Gen 4.5 video model to the test. He focused on physical accuracy and prompt adherence.
- Physics: He showed a glass sphere rolling down marble stairs with water interactions. The visual physics were incredibly realistic.
- Consistency: The model handled complex lighting and character movements well, though the creator felt the “anime style” test was a bit weak.
- The Catch: He pointed out that it still lacks audio generation, which puts it slightly behind competitors that generate sound effects automatically.
🏎️ Rivian and Open Source
This savvy professional also covered Mistral’s new “Devstrol 2” model, which brings powerful coding capabilities to local machines. On the hardware side, the host attended a Rivian event, revealing their roadmap for “eyes-off” autonomous driving and a new voice assistant that controls physical car functions.
📉 The “Slop” Fatigue
Finally, the expert touched on the backlash against a recent AI-generated ad from McDonald’s. He explained that audiences are feeling fatigued by low-effort AI visuals from billion-dollar companies that could afford real actors.
📊 Stats of the Day
Here are the specific numbers the expert shared regarding the new GPT-5.2 model:
- Context Window: 400,000 tokens (approx. 300k words).
- Max Output: 128,000 tokens.
- Input Cost: $1.75 per million tokens.
- Output Cost: $14.00 per million tokens.
If you want to see the video tests of the physics engine or the full breakdown of the benchmarks, you should definitely watch the full breakdown linked below!