- AI Valley
- Posts
- AI in 2025
Together with
Howdy, it’s Barsee.
Happy Wednesday, AI family, and welcome to another AI Valley edition.
Today’s climb through the Valley reveals:
OpenAI said to develop “Garlic” model
Anthropic reportedly preparing for one of the largest IPOs ever
What to be thankful for in AI in 2025
Plus trending AI tools, posts, and resources
Let’s dive into the Valley of AI…
REMIO 2.0
Image Credit: Remio
Tired of scattered notes, lost files, and forgotten conversations? Remio unifies your digital life, transforming your knowledge into a personal intelligence source.
Here’s what you’ll love about remio:
Tailored Answers: remio provides unique answers by combining with your personal knowledge.
No More Uploads to ChatGPTs: One click to sync all your files, making your entire knowledge base chatable with AI.
Master Your Meeting: Unlimited free recording with transcription, get AI summaries with key decisions.
Privacy & Security: With a "Local First" design, all your data is stored exclusively on your device.
*This is sponsored
THROUGH THE VALLEY
Source: The Information
OpenAI is reportedly working on a new model called “Garlic” after internally calling a Code Red in response to Google’s recent Gemini gains. Chief Research Officer Mark Chen briefed staff on the project, describing Garlic as a smaller, more efficient model that still matches or beats some Gemini 3 results in internal tests. This model addresses structural issues found in earlier versions like GPT-4.5 and could be released as GPT-5.2 or 5.5 early next year.
Image Credit: Getty Images
Anthropic is in early talks to go public next year. The company has hired Wilson Sonsini, the law firm behind major tech IPOs like Google and LinkedIn, and is also raising a private round that could value it above $300 billion, backed by Microsoft and Nvidia. Discussions with banks are still early, and Anthropic says no decision has been made. The talks come as the startup expands quickly with a $50 billion data center plan and recent hires tied to Airbnb’s IPO, putting it in a potential race with OpenAI for a public listing.
PEAK OF THE DAY
1/ OpenAI kept a fast release pace in 2025
Source: OpenAI
GPT-5 launched in August, followed by GPT-5.1 in November with Instant and Thinking modes that adjust how much time the model spends on a task. The release began with math and coding issues, but fixes came quickly. Companies using GPT-5 are seeing strong results. ZenDesk reports the model now resolves more than half of customer tickets, with some clients reaching 80 to 90 percent resolution.
On March 25, OpenAI released 4o image generation, which went viral all over the world with Ghibli image styles. Developers also got a major upgrade with GPT-5.1-Codex-Max, a coding model built for long, multi-step workflows. ChatGPT Atlas pushed the assistant directly into the browser, adding summaries, on-page analysis, and search into normal browsing. Sora 2 expanded OpenAI’s video work with better physics, synced audio, and more creative control, along with a new Sora app for sharing videos.
OpenAI also released gpt-oss-120B and gpt-oss-20B, its first serious open-weight models since GPT-2. Early feedback is mixed, but it marks a major shift in how the company approaches open development.
2/ China’s open-weight AI ecosystem went fully mainstream in 2025
Source: a16z
A study from MIT and Hugging Face found that China now slightly leads the U.S. in global open-model downloads, driven mainly by DeepSeek and Alibaba’s Qwen family. DeepSeek-R1 set the tone early in the year with MIT-licensed reasoning weights and a full set of distilled variants. Moonshot followed with Kimi K2 Thinking, a tool-using reasoning model seen as one of the strongest open options available. Z.ai released GLM-4.5 and GLM-4.5-Air as agent-focused models, while Baidu open-sourced its ERNIE 4.5 MoE suite under Apache 2.0. Alibaba kept shipping throughout the year with its Qwen3 line, including coding, reasoning, and multimodal models, earning the nickname “Qwen’s summer.” For anyone relying on open ecosystems or on-prem setups, 2025 was the year China’s open-weight scene became a serious global contender.
3/ Google answered GPT-5 with Gemini 3 and Nano Banana Pro
Credit: sid
Gemini 3 is its most capable model so far, built for stronger reasoning, coding, and multimodal work. It also introduced a Deep Think mode for slow, complex problems. The unexpected standout, though, is Nano Banana Pro, Google’s new image generator inside Gemini 3 Pro. It excels at infographics, diagrams, multi-subject scenes, and clear multilingual text at high resolutions. For enterprise teams that rely on charts, product visuals, and system diagrams, this kind of accuracy matters far more than creative art outputs.
TRENDING TOOLS
Cua App Use > Create virtual desktops for AI agents to focus on specific apps
Bloom > World's first on-brand AI
Virlo AI > Spot 3 breakout trends on tiktok before they become hit
LegalRobot > AI contract analysis tool that helps you understand complex legal language and flags potential risks in documents
THINK PIECES / BRAIN BOOST
THE VALLEY GEMS
What’s trending on social today:
The number one song on Billboard’s Country Digital Song Sales chart is AI-generated.
THAT’S ALL FOR TODAY
Thank you for reading today’s edition. That’s all for today’s issue.
💡 Help me get better and suggest new ideas at [email protected] or @heyBarsee
👍️ New reader? Subscribe here
Thanks for being here.
HOW WAS TODAY'S NEWSLETTER |
REACH 100K+ READERS
Acquire new customers and drive revenue by partnering with us
Sponsor AI Valley and reach over 100,000+ entrepreneurs, founders, software engineers, investors, etc.
If you’re interested in sponsoring us, email [email protected] with the subject “AI Valley Ads”.