Microsoft is gearing up for GPT-5

Welcome back.

Microsoft doesn’t want a repeat of last year’s GPT-4o surprise.

This time, they’re gearing up early, preparing for OpenAI’s upcoming GPT-5, and it looks like they’re betting big on this one.

Let’s dive in.

In today’s release:

1. Microsoft prepares for GPT-5

2. Google Gemini’s upgrade

3. DeepSeek to open-source AI models and more

Microsoft prepares servers for OpenAI’s GPT-5 launch

Microsoft is gearing up its server infrastructure to support OpenAI’s upcoming GPT-4.5 and GPT-5 models, according to a report from The Verge.

This time, the tech giant is taking no chances after being caught off guard by GPT-4o’s release last year, which unexpectedly undermined Azure’s paid AI services like speech and translation.

• Launch timeline: GPT-4.5 is expected in the coming weeks, with GPT-5 likely arriving in late May.

• Unified model strategy: GPT-5 may integrate the o3-mini model, merging GPT and GPT-o lines for simplicity.

• Orion codename: GPT-4.5, dubbed “Orion,” will be OpenAI’s last non-chain-of-thought model before shifting strategies.

Sam Altman has hinted at a unified approach for OpenAI’s future models, aiming to eliminate the need for users to pick between model types. GPT-5 is expected to reflect this strategy, combining the strengths of GPT and o-series models into a single, more intuitive experience.

How can you take advantage of this?

With GPT-5’s unified intelligence approach, users can expect fewer limitations and better results without juggling multiple models. This could streamline workflows for developers, businesses, and individuals using AI for diverse tasks, from content generation to complex data analysis.

Google adds a file analysis feature to Gemini

Google is rolling out an upgrade to Gemini’s non-paid tier, adding file upload and analysis capabilities — a feature previously reserved for Gemini Advanced subscribers.

Users can now upload documents, code files, and spreadsheets to Gemini for summarization, analysis, or even chart creation directly from the chatbot.

• File support: Uploads include PDFs, Word docs, Google Sheets, CSVs, code files (C, Java, PHP, HTML), and more.

• Multi-platform rollout: Available on the web and Android (iOS support for Drive uploads is still pending).

• Limitation: Up to 10 files at once, each being up to 100MB in size, similar to Gemini Advanced.

This move helps Gemini catch up with ChatGPT, which introduced file uploads for non-paying users back in May 2024. While there are still some limitations, like not being able to upload a file and image in the same prompt, this update makes Gemini’s offering more competitive.

How can you take advantage of this?

You can now use Gemini to quickly summarize large documents, analyze spreadsheets, or review code without upgrading to a paid plan. It’s an efficient way to save time on research, document reviews, and data analysis, without spending a dime.

DeepSeek to open-source AI models, code, and data

Chinese AI startup DeepSeek is taking a bold step toward transparency by open-sourcing key components of its AI technology next week.

The company plans to release not just its AI models, like the well-regarded R1, but also the underlying code, datasets, and development processes going further than its rivals like OpenAI and Meta.

• What’s coming: Five code repositories, including R1’s code, data, and documentation, will be made public.

• Open-source focus: DeepSeek aims to surpass Meta’s and OpenAI’s limited open-source efforts.

• Security concerns: Governments are already expressing caution about the broader access to advanced AI models.

DeepSeek’s open-source approach not only intensifies competition in the global AI race but also challenges the norms of keeping proprietary AI tech under wraps. Its surprising advancements, outpacing some of OpenAI’s and Meta’s models, have forced competitors like Baidu to rethink their strategies.

How can you take advantage of this?

Developers and researchers will soon gain access to DeepSeek’s powerful AI models and training data, enabling them to build, customize, or improve advanced AI tools without the usual barriers. This could lower costs and accelerate innovation for businesses and individuals.

OTHER AI NEWS

ChatGPT Pro users can now use Operator: Operator is live for users in Australia, Brazil, Canada, India, Japan, Singapore, South Korea, the UK, and most places where ChatGPT is available. EU availability is still in progress.

AI is coming to iPhone 15 Pro: Apple’s Visual Intelligence feature is coming to iPhone 15 Pro via a future update. It’ll work through the Control Center or Action button, similar to the new iPhone 16e.

Google’s AI for job seekers: Google’s Career Dreamer tool uses AI to match skills with careers and draft resumes. It aims to simplify job searching and boost your chances.

OpenAI’s revenue goal in 2025: OpenAI’s CFO says tripling revenue to $11B next year is “definitely possible.” User growth is up 33%, fueling talks of a SoftBank investment.

AI recruiting startup’s staggering valuation: Mercor, founded by 21-year-old Thiel Fellows, raised $ 100M, valuing it at $ 2B. Its AI platform streamlines hiring, boasting $ 75M ARR and partnerships with top AI labs.

TOP AI TOOLS THIS WEEK

Chance AI for iOS: AI-powered visual search

BookWatch: Animated book summaries for visual learners

Kraftful 4.0: The AI-native way to build products

HeyGen iOS: Text to lifelike videos, your avatars, anytime, no editing

TensorPool: The easiest way to use cloud GPUs

AND THAT’S A WRAP

Thank you for reading!

If you found this email useful, share it with a friend or colleague who also loves AI.

Also, drop me a follow on Twitter/X for more AI and tech updates.

I’ll talk to you soon.

Mike