Welcome back!

OpenAI isn’t slowing down. With the launch of o3-mini, it’s advancing what’s possible for both free and paid users. This new reasoning model brings faster performance and affordable pricing that undercuts its competitors like DeepSeek. Add in a sneak peek into the future of GPT-5 and you've got a crazy week. Let’s unpack...

Today’s Summary:

• 🚀 OpenAI launches o3-mini

• 🤖 OpenAI roadmap: GPT-5 & beyond

• 📖 US rules on AI art copyright

• 🌊 Mistral Small 3 rivals GPT-4o mini

• 📈 Alibaba Qwen 2.5-Max outperforms DeepSeek

• 📹 ChatGPT Voice Mode adds video input in the EU

• 🛠️ 2 new tools

TOP STORY

OpenAI launches o3-mini for all users

The Summary: OpenAI released o3-mini, a new AI model that matches its bigger cousin o1 in math and coding tasks while running 24% faster. The model brings advanced reasoning to ChatGPT's free tier for the first time, while paid users get access to a higher-performance "high" mode that can match or beat o1 on several benchmarks. All versions include a new web search feature and affordable API pricing.

Key details:

• Three modes: low (fast), medium (balanced), high (max accuracy)

• Free users can access via "Reason" button (rate limits apply)

• Plus/Team users get 150 messages per day (3x from o1-mini 50)

• Pro users get unlimited access to o3-mini and o3-mini-high

• Model processes 200K input tokens and 100K output tokens, beating GPT-4o's 128K/16K

• API pricing at $0.55 / million input tokens makes it 63% cheaper than o1-mini, matching DeepSeek R1 pricing

• AIME math test: 83.6% accuracy in high mode (beats o1)

• Competitive coding: 2130 ELO score (vs o1's 1891) at 1/10th the cost

• Web search included

Why it matters: The race between US and Chinese AI is driving innovation while bringing costs down - a win for users worldwide. o3-mini reveals two major market shifts: first, AI reasoning is becoming a commodity - the 90% cost reduction while matching o1's performance shows the marginal cost of "intelligence" is dropping faster than predicted. Second, OpenAI's aggressive push into specialized models (like STEM-focused o3-mini) rather than general-purpose ones hints at the limits of the "single model for everything" approach.

o3-mini in ChatGPT - FAQ

OPENAI

OpenAI reveals what’s next for GPT-5, automation, and robots

The Summary: In a Reddit AMA, OpenAI CEO Sam Altman and his team shared new updates on the company’s roadmap, covering everything from GPT-5 development to robotics and advanced features. Altman confirmed work on expanding context length, tool integration, and model capabilities while addressing questions on pricing, open-source, and the company's focus on scaling compute power through projects like Stargate.

Key details:

• GPT-5 development is confirmed. Altman hinted at dropping the "5o" label, sticking with "GPT-5". No set timeline.

• Tool unification will allow users to seamlessly switch between speaking, coding, and document editing

• The robotics division's first goal is a small production run of highly capable robots. The focus is on real-world learning over mass production

• OpenAI aims to lower Plus plan costs over time

• File attachment support is in development for o3

• Context length expansion is underway

• OpenAI plans to show more detail of the o1/o3 "internal thinking"

• Memory support for "o" series is in progress

• OpenAI is still working on GPT-4o improvements

• Stargate is OpenAI’s "compute factory" for scaling AI models

• OpenAI’s approach to open-source is still under discussion internally, with Altman admitting the company may have been "on the wrong side of history" regarding open-source

• The company is exploring the integration of advanced automation features beyond the browser, including full computer automation

• OpenAI is considering adding Operator to the regular Plus plan

Why it matters: OpenAI revealed a roadmap filled with practical enhancements aimed at expanding AI capabilities, improving user experience, and maintaining competitive momentum in a rapidly evolving landscape.

AI COPYRIGHT

Copyright Office rules on AI art protection

The Summary: The US Copyright Office released new AI guidelines reaffirming that AI-generated art created solely from text prompts does not qualify for copyright protection. However, works incorporating AI with clear human creative input may still be protected. The Office maintains that existing copyright laws apply, but require a case-by-case evaluation of human contributions to AI-generated content.

Key details:

• AI prompts, no matter how detailed, do not qualify as authorship, similar to giving instructions rather than creating

• Selecting a single AI-generated output from a variety of outputs isn’t authorship, comparable to curating a living garden rather than painting

• AI-assisted works may be copyrighted if they involve sufficient human input, selection, coordination, and creative modification

• The Office will release a follow-up report later this year on AI-related copyright infringement and fair use

Why it matters: The guidance reinforces human authorship as the foundation of copyright, preventing purely AI-generated works from receiving the same protections. Artists using AI as a tool can still claim protection, though the exact threshold of required human input remains unclear.

Read the full report

QUICK NEWS

Quick news

• Mistral Small 3: open-source replacement for GPT-4o mini

• Alibaba Qwen 2.5-Max tops DeepSeek-V3

• ChatGPT Voice Mode with video input is now available in the EU

TOOLS

🥇 New tools

• Stella - Automate your daily admin tasks with AI agents

• QwenChat - Alibaba chatbot, including Qwen 2.5-Max frontier model