AI NEWS: GPT-5.2 Is HERE! (Plus Runway 4.5 and the Next Image AI Leaks)

AI NEWS: GPT-5.2 Is HERE! (Plus Runway 4.5 and the Next Image AI Leaks)

by Matt Wolfe

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 Matt Wolfe 👥 1.0M 📅 December 12, 2025 ⏱ 34 min 👁 53K 📄 news review 🧭 2026-08-28
Available in: English (current) Français

Keywords

GPT-5.2Runway Gen-4.5OpenAIAI agentsAI news

Summary

In this weekly AI news update, Matt Wolfe covers the latest developments in the AI industry. The main story is the release of OpenAI’s GPT-5.2 model, which shows improvements in math and science benchmarks, with a 400k token context window and competitive pricing. The video also discusses a leaked new OpenAI image model, potentially named Chestnut and Hazelnut, which appears to generate realistic celebrity images and handle text in images. Additionally, OpenAI announced a partnership with Disney, investing $1 billion, and the formation of the Agentic AI Foundation with Anthropic and Block to standardize AI agents. The host then provides a hands-on review of Runway’s Gen-4.5 video model, testing its physical accuracy, character emotion, animation style, macro details, and cinematic action, concluding it excels in prompt adherence but lacks audio. The video also covers new open-source models from Mistral (Devstral 2), Zhipu AI (GLM 4.6V), and Alibaba (Qwen3 Omni Flash), along with rapid-fire news including OpenAI removing ad-like suggestions, ChatGPT integrating Adobe apps, Meta acquiring the Limitless Pendant, and the backlash against McDonald’s AI-generated Christmas ad. The host ends with a commentary on ‘AI slop’ fatigue and the importance of quality in AI-generated content.

194 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides valuable, up-to-date information on AI developments, with a focus on practical implications. The host’s hands-on testing of Runway Gen-4.5 adds concrete evidence to the discussion, and his balanced commentary on the strengths and weaknesses of the model is informative. The argumentation is generally solid, with clear explanations of technical concepts like token context and benchmarks. However, some claims, such as the leaked image model, are based on speculation and lack verification. The host also occasionally makes subjective judgments without deep analysis, but overall the information is presented in an accessible and engaging manner.

Scientific Rigor, Source Quality, Title Accuracy

The video references official sources for major announcements, such as OpenAI’s blog posts and Mistral’s news page, and includes links in the description. The host also cites specific benchmarks and provides context for comparisons. However, some information is based on leaks and unverified reports, which are presented as such. The title accurately reflects the content, and the video’s structure with chapters helps viewers navigate. The host’s personal testing adds credibility, but the lack of independent verification for some claims slightly reduces the overall rigor.

195 words

Title / Content Match

The title accurately reflects the content, highlighting the main topics: GPT-5.2, Runway 4.5, and image model leaks.

Quality & Reliability

7/10

The video is a weekly AI news roundup with practical demonstrations and references to official sources. The host provides balanced commentary, acknowledges uncertainties, and includes hands-on testing of Runway Gen-4.5. However, some claims rely on leaks and unverified benchmarks, and the presentation is informal.

Chapters

Cited Sources

  • Introducing GPT-5.2 — Official OpenAI announcement of GPT-5.2, detailing its capabilities and benchmarks.
  • Agentic AI Foundation — OpenAI's announcement of the Agentic AI Foundation, a collaboration to standardize AI agents.
  • Devstral 2 and Mistral Vibe CLI — Mistral's announcement of Devstral 2, an open-weight coding model, and the Vibe CLI tool.
  • Zhipu AI's GLM-4.6V — VentureBeat article covering Zhipu AI's release of GLM-4.6V, an open-source vision-language model with tool calling.
  • McDonald's AI-generated Christmas ad — The Verge article discussing the backlash against McDonald's AI-generated Christmas ad.
  • ChatGPT apps: Adobe Photoshop, Acrobat Express — The Verge article about ChatGPT integrating Adobe apps for photo and PDF editing.

Concurring Sources

  • OpenAI's GPT-5.2 announcement — Official source confirming the release and specifications of GPT-5.2.
  • Agentic AI Foundation announcement — Official source confirming the formation of the Agentic AI Foundation.
  • Mistral's Devstral 2 release — Official source confirming the release of Devstral 2 and Vibe CLI.

Dissenting Sources

  • Leaked OpenAI image model — The video discusses leaked images and speculation about a new OpenAI image model, but no official source is provided, and the information is unverified.

External References

Contribution & Novelties

The video provides a timely and comprehensive overview of the latest AI news, with a unique hands-on testing segment for Runway Gen-4.5. The host’s practical demonstrations offer insights into the model’s capabilities and limitations, which is valuable for viewers considering using such tools. The discussion of the Agentic AI Foundation highlights an important industry trend towards standardization. The video also touches on the societal implications of AI-generated content, such as the ‘AI slop’ phenomenon.

Pour aller plus loin :

  • Agentic AI Foundation — Official announcement of the foundation, explaining its goals and members.
  • Runway Gen-4.5 — Runway’s official page for their video generation models, including Gen-4.5.
  • Mistral Devstral 2 — Mistral’s announcement of Devstral 2, an open-weight coding model.
  • AI slop — Wikipedia article on the term ‘AI slop’, referring to low-quality AI-generated content.

134 words

Radar Profile

The radar profile shows a balanced performance across all dimensions, with slightly higher scores in information quantity and technical level, reflecting the video's comprehensive coverage and hands-on testing. The lower score in information quality is due to reliance on some unverified leaks and subjective assessments.

Reliability 7/10

💬 Très positif. Sur les 30 commentaires analysés, la majorité exprime une appréciation chaleureuse pour le contenu et la présentation de Matt Wolfe, avec des éloges pour sa clarté et son enthousiasme, et un intérêt marqué pour les sujets couverts comme GPT-5.2 et Runway Gen-4.5.