AI Valley
Posts
A new OpenAI model coming next week

A new OpenAI model coming next week

PLUS: New DeepSeek releases two new models

Barsee
02 Dec

_{Sign up}_|_{Follow us on X}_|_Sponsor

Together with

Howdy, it’s Barsee.

Happy Tuesday, AI family, and welcome to another AI Valley edition.

Today’s climb through the Valley reveals:

A new OpenAI model coming next week, outperforming Gemini 3
Runway unveils Gen-4.5 video generation model that tops Video Arena
DeepSeek releases two new models
Plus trending AI tools, posts, and resources

Let’s dive into the Valley of AI…

OUTSKILL

Still doing manual tasks repetitively? Learn AI to automate 50% of your work

Image Credit: Outskill

It is almost 2026, and if you have been meaning to learn AI but have not found the time, this could be a good opportunity to start.

Outskill is hosting a 2-day live AI Mastermind that focuses on practical skills including AI tools, automations, and building simple agents. The goal is to help participants understand how to use these systems more effectively in real work.

They are running a holiday giveaway where the first 100 participants can join at no cost. The regular price is 395 dollars.

🧠Live sessions- Saturday and Sunday
🕜10 AM EST to 7PM EST

_{*This is sponsored}

THROUGH THE VALLEY

1/ A new OpenAI model coming next week, outperforming Gemini 3

Source: The Information

OpenAI is preparing to launch a new reasoning model that Sam Altman says outperforms Google’s upcoming Gemini 3 in internal tests, with a release planned as early as next week. The push is part of an internal “Code Red” effort as Google’s Gemini user base climbs and its models regain benchmark leadership. To focus on the launch, OpenAI is pausing several projects, including advertising plans, new autonomous agents, and its “Pulse” briefing service. The company is shifting resources toward core ChatGPT upgrades and urgent improvements in image generation.

2/ Runway unveils Gen-4.5 video generation model that tops Video Arena

Source: Runway

Runway has released Gen-4.5, a new AI video model that now sits at the top of Artificial Analysis’ text-to-video leaderboard. The company says it produces more natural motion, better physics, and stronger realism, with details like hair and fabric staying consistent across frames. It handles many styles but is strongest in cinematic and photorealistic shots. Gen-4.5 is rolling out to all users in the coming days and includes upgrades in training quality, control modes, and GPU performance. Runway says it’s the closest step yet toward true film-grade AI video.

3/ DeepSeek releases two new models

Source: DeepSeek

DeepSeek has released two new versions of its experimental AI model, now called DeepSeek-V3.2, with major upgrades in reasoning and tool use. The company says the model performs on par with GPT-5 on several reasoning tests and can use tools like search, calculators, and code runners directly during its thinking process, and it is almost 30x cheaper than Gemini 3.0 pro. A second version, V3.2-Speciale, targets advanced math and long-form problem solving, matching results from Google’s Gemini-3 Pro and scoring at Olympiad-level on math and programming exams. The releases continue DeepSeek’s push to advance open-source AI and strengthen its lead in China’s fast-moving AI race.

TRENDING TOOLS

Join MongoDB.local San Francisco on Jan 15 where builders and innovators shape what’s next in AI *
Kling O1 > this new model lets you edit videos with prompts and images – like Nano Banana but for video
Whisper Thunder > Gen-4.5, a new video generation model from Runway that beats OpenAI and Google in key benchmarks
Unfold > It turns any topic into interactive, AI-powered courses. Powered by Gemini 3, it adapts to your learning style
Agenta > Build reliable LLM apps together with integrated prompt management, evaluation, and observability