# 📆 ThursdAI - Sep 11 - SeeDream 4, Lucy 14B, ChatGPT gets MCP, OpenAI $300B deal with Oracle, Qwen...

Source: https://www.youtube.com/watch?v=tWJIWFDzxFc
Recap page: https://rapidrecap.app/video/tWJIWFDzxFc
Generated: 2025-09-12T03:33:02.387+00:00

---
## Quick Overview

The ThursdAI news roundup for September 11th covered significant AI developments including the release of SeeDream 4 for rapid image generation, the introduction of Lucy 14B with impressive image-to-video speeds, OpenAI's potential $300 billion deal with Oracle for compute resources, the integration of MCP support into ChatGPT, and updates to open-source models like Qwen and Kimi.

**Key Points:**
- OpenAI is reportedly in talks for a $300 billion deal with Oracle, which would secure substantial compute resources for OpenAI, with Larry Ellison's wealth significantly increasing due to this potential partnership.
- New image generation models are rapidly advancing, with SeeDream 4 capable of generating 2K images in 1.8 seconds and Lucy 14B offering fast image-to-video generation, both pushing the boundaries of speed and quality.
- ChatGPT now supports full MCP (Model Connectors and Plugins) integration through developer mode, leveraging its 400-500 million weekly users to access a vast array of tools.
- Open-source LLM updates include Qwen 3 ASR Flash for multilingual speech recognition, Kimi K2 0905 doubling its context length to 256K for enhanced agentic and coding tasks, and Tencent's Hunion image 2.1 offering 2K resolution text-to-image generation with enhanced prompt capabilities.
- Mistral secured a $1.3 billion investment at a $13.8 billion valuation, highlighting its growth and importance in the European AI landscape, with ASML being a key investor.
- New stealth coding models, Sonoma Dusk and Sonoma Sky, emerged with a 2 million token context window, showing high speed and quality for code generation.

**Context:** This episode of ThursdAI, hosted by Alex Walov, features co-hosts Wolf from Riven, Yam, LDJ, and Niston discussing the latest AI news for September 11th. The panel highlights a mix of major industry announcements, new model releases, and open-source advancements, covering developments in LLMs, image generation, video synthesis, and audio processing.

## Detailed Analysis

The ThursdAI September 11th episode provided a comprehensive overview of the week's AI news. Key highlights included the rapid advancement in image and video generation models, such as Bytedance's SeeDream 4, which generates 2K images in 1.8 seconds, and the cart's Lucy 14B, offering fast image-to-video capabilities. OpenAI's significant potential $300 billion deal with Oracle for compute resources was a major financial announcement, impacting Oracle's stock and Larry Ellison's net worth. ChatGPT's integration of full MCP support was lauded as a major step in expanding its utility for its hundreds of millions of users. The open-source landscape saw updates to models like Qwen, with Qwen 3 ASR Flash improving multilingual speech recognition, and Kimi K2 0905 doubling its context length to 256K, enhancing its agentic and coding performance. Tencent released Hunion image 2.1, an open-source text-to-image model with 2K resolution and advanced prompt handling. New stealth coding models, Sonoma Dusk and Sonoma Sky, emerged with an impressive 2 million token context window. Mistral's growth was underscored by a $1.3 billion investment, solidifying its position as a major player. The discussion also touched upon Claude's new file creation/editing capabilities and potential bugs, as well as the emergence of tools like Ebisense for video editing.

### AI News Roundup

- SeeDream 4 release for rapid image generation
- Lucy 14B for fast image-to-video
- OpenAI's reported $300B Oracle deal
- ChatGPT adds full MCP support
- Qwen, Kimi, and Hunion model updates
- Sonoma Dusk/Sky coding models debut
- Mistral's $1.3B investment
- Claude's code interpreter feature

### Key Model Releases

- SeeDream 4 generating 2K images in 1.8s
- Lucy 14B image-to-video model
- Hunion image 2.1 with 2K resolution and long prompts
- Sonoma Dusk and Sky with 2 million context window

### Major Partnerships & Investments

- OpenAI's potential $300 billion deal with Oracle for compute
- Mistral secures $1.3 billion investment at $13.8 billion valuation from ASML

### ChatGPT Enhancements

- Full MCP support integrated via developer mode
- Expansion of tool access for 400-500 million weekly users

### Open Source Updates

- Qwen 3 ASR Flash for multilingual speech recognition
- Kimi K2 0905 doubles context to 256K
- Tencent Hunion image 2.1 released
- Ernie 4.5 open-source model available

### Visual & Audio AI Developments

- SeeDream 4's speed and quality in image generation
- Lucy 14B's rapid image-to-video capabilities
- Qwen 3 ASR Flash for speech recognition
- Stable Diffusion's audio 2.5 update

### LLM & Agent Advancements

- Sonoma Dusk/Sky models with 2 million context window
- Kimi K2 0905's extended context for agentic use
- Salesforce fine-tunes deep research model for multi-tool usage