- Outlier Coder's Newsletter
- Posts
- Coder's Newsletter # 5
Coder's Newsletter # 5
🤖 Customized ChatGPT for Every Country? OpenAI Thinks So
Hey Outliers,
Welcome back to Outlier coder’s Newsletter! 🎉 Your front-row seat to AI’s biggest moves, smarter tools, and key research updates.
From DIY search training to budget-friendly reasoning and infrastructure that keeps AI closer to home, this week’s updates ask a simple question: what happens when you stop trying to build the biggest model—and start building the right one?
This Week in AI
Alibaba trains LLMs to search without Google
Mistral’s new model brings power without the price tag
OpenAI partners with nations to build secure, localized AI
🔥 What’s hot this week?🔥
🔍 Alibaba’s ZEROSEARCH: Training LLMs to Search

Alibaba has introduced ZEROSEARCH, a new reinforcement learning framework that teaches large language models (LLMs) to search for information-without ever accessing real search engines. This approach aims to boost LLMs’ reasoning and retrieval skills while dramatically cutting training costs.
Key Highlights:
ZEROSEARCH fine-tunes an LLM to act as a mock search engine, generating both helpful and noisy documents for any query.
A curriculum schedule gradually makes retrieval tasks harder, so the model’s reasoning skills improve step by step.
By skipping commercial search APIs, training costs drop by up to 88%, and you stay in control of the quality of documents used.
In tests, models trained with ZEROSEARCH often match or outperform those trained with real search data, especially when larger LLMs are used for the mock search.
The method works with most LLM sizes and reinforcement learning algorithms; the main extra cost is GPU time for running your in-house search simulator.
You avoid unpredictable web results and can ensure your model only sees data you approve.
What This Means for You:
ZEROSEARCH lets you train AI models that act like they have “Google instincts,” but without paying for every search or worrying about what’s out there on the web. You control all the training data, so your AI is safer and more reliable.
This means you can build smart chatbots, summarizers, or Q&A tools cheaply and confidently-without surprise bills or messy web content sneaking in.
🤖 Mistral Medium 3: Enterprise AI Gets Smarter and Cheaper
Introducing Mistral Medium 3: our new multimodal model offering SOTA performance at 8X lower cost.
- A new class of models that balances performance, cost, and deployability.
- High performance in coding and function-calling.
- Full enterprise capabilities, including hybrid or— Mistral AI (@MistralAI)
2:12 PM • May 7, 2025
Mistral AI has just unveiled Medium 3, a next-generation AI model designed for professional use. It’s making waves for delivering nearly top-tier performance in coding, STEM, and multimodal tasks-at a price point that undercuts the competition.
Key Highlights:
Mistral Medium 3 delivers over 90% of the performance of leaders like Claude 3.7 Sonnet, making it a strong pick for enterprise-grade use.
It excels in coding, science, and math—ideal for technical teams tackling automation, analysis, and complex problem-solving.
With support for both text and images, it handles multimodal tasks like document parsing and visual data analysis with ease.
Token costs are up to 8x lower than comparable models, making large-scale deployment far more cost-effective.
It’s resource-efficient too—runs on just 4 H100 GPUs, whether on-prem or in the cloud.
Supports 40+ languages and custom fine-tuning, so teams can adapt it to their exact needs.
What This Means for You:
Mistral Medium 3 delivers near-top-tier AI at much lower cost, enabling budget-friendly large-scale projects. It runs efficiently on just four GPUs, so deployment is fast and hardware needs are minimal.
This means smarter code reviews, quicker data analysis, and responses tailored to the way your team works-like having an expert who speaks your internal language.
🌍 OpenAI for Countries: Bringing Stargate to Nations Worldwide

OpenAI has launched “OpenAI for Countries,” an ambitious new initiative under its Stargate project. The goal: help countries build their own secure, sovereign AI infrastructure-rooted in democratic values-so they can harness AI’s benefits while maintaining control over their data and future.
Key Highlights:
OpenAI will partner with countries to build local AI data centers, ensuring sensitive data stays secure and within national borders.
Each country receives a customized ChatGPT tailored to its languages, culture, and specific public service needs.
The initiative focuses on “democratic AI” that upholds individual freedoms and prevents government overreach.
Countries will also co-invest in national AI startup funds, seeding new local industries, jobs, and innovation ecosystems.
The first phase aims for 10 partnerships with individual countries or regions, with plans to expand further.
What This Means for you:
As an AI user, OpenAI for Countries means you’ll get access to smarter, more locally relevant AI tools-like customized ChatGPT-tailored to your language, culture, and privacy needs. Everyday tasks, from getting health advice to navigating government services, could become easier and more efficient thanks to AI that truly understands your context.
If you work with technology, this initiative offers access to powerful local AI infrastructure and funding, letting you build, deploy, and secure solutions that truly fit your region’s needs.
☄️ Trending Bytes
You either die a nonprofit, or live long enough to become a… wait what is OpenAI again? 🫠
🗓️ Save the Date: Webinar on May 15— Sign Up Today!
Find out why traditional benchmarks fall short—and what’s next for AI evaluation.
Click ⬇️ to signup
💡 AI Model Spotlight
Model Name | Parent Company | Release Date | Key Highlights |
---|---|---|---|
Mistral AI | May 7, 2025 | 90% of Claude Sonnet 3.7 performance, Cost-effective deployment, Open-source availability | |
Figma | May 7, 2025 | Claude Sonnet 3.7 integration, Modular AI tools, Focus on creative workflows | |
May 6, 2025 | Superior coding performance, Advanced multimodal reasoning, Early access release | ||
IBM | May 6, 2025 | Wide range of AI agents, Focus on interoperability, Support for various industries | |
AI2 | May 1, 2025 | Efficient performance, Open-source release, Suitable for resource-constrained environment |
📋 Feedback
How would you describe your experience with this edition of the newsletter? |
That’s it for now—keep pushing AI forward! 🚀
You received this email because you are subscribed to Outlier.ai. The content of this email is for informational purposes only and may not be reproduced or distributed without written permission. AI research is rapidly evolving, and while we strive for accuracy, we encourage readers to verify details from official sources.
Please note that all emails exchanged with Outlier.ai may be subject to monitoring for compliance with company policies. Information contained in this email is confidential and intended solely for the recipient. No legally binding commitments are created by this email.
All trademarks used in this email are the property of their respective owners. You are receiving this email because you have authorized Outlier.ai to send you updates. For more details, visit the Outlier.ai website. Terms & Conditions apply.