OpenAI Drops the Most Advanced AI

Welcome to another edition of AI PlanetX.

Breakthroughs from OpenAI, Google & Microsoft: o3/o4 released, Gemini 2.5 Flash slashes costs, CPU-based AI model emerges.

• Hottest AI News

• Top AI & SaaS Tools

• Top AI & Tech News

• Interesting Uses of AI: Prompt That Makes AI Think Like a Team

• AI Video Tutorial

• F-R-E-E AI Course of the Day: Building AI Browser Agents by Co-Founders of AGI

Hottest AI News

OpenAI

OpenAI Launches o3 and o4 With Unprecedented Reasoning Capabilities

OpenAI has introduced two reasoning models, o3 and o4-mini, which pause to methodically process questions before answering, marking a major advancement in AI.

Details:

• o3 is OpenAI's most advanced model, excelling in math, coding, science, reasoning, and visual tasks. o4-mini offers a cost-effective, fast, and balanced solution for developers, integrating with tools like web browsing, Python, and image processing

• These models "think with images," analyzing visuals like sketches or diagrams, even blurry ones. They run Python code in browsers and can search the web for real-time info

• Available to Pro, Plus, and Team subscribers, these models are also offered via API: o3 at 10/40 per million tokens, o4-mini at 1.10/4.40. o3-pro will be released soon for ChatGPT Pro users

This release boosts OpenAI’s edge over rivals like Google and Meta. CEO Sam Altman suggests o3 and o4-mini could be the last standalone reasoning models before GPT-5 combines all capabilities.

You’ve heard the hype. Now it’s time for results

After two years of siloed experiments, proofs of concept that fail to scale, and disappointing ROI, most enterprises are stuck. AI isn't transforming their organizations — it’s adding complexity, friction, and frustration.

But Writer customers are seeing a positive impact across their companies. Our end-to-end approach is delivering adoption and ROI at scale. Now, we’re applying that same platform and technology to bring agentic AI to the enterprise.

This isn’t just another hype train that doesn’t deliver. The AI you were promised is finally here — and it’s going to change the way enterprises operate.

See real agentic workflows in action, hear success stories from our beta testers, and learn how to align your IT and business teams.

Watch the product showcase now

DeepMind

Google Gemini 2.5 Flash Slashes AI Costs With Thinking Budgets

Google has launched Gemini 2.5 Flash, an AI upgrade that helps businesses cut costs by up to 600% by controlling AI’s computational workload. This move balances advanced reasoning with competitive pricing in the crowded AI market.

Details:

• The "thinking budget" lets developers control AI power, with costs from $0.60 to $3.50 per million tokens. This creates Google’s "hybrid reasoning model," adjusting from 0 to 24,576 tokens based on task complexity

• Gemini 2.5 Flash scores 12.1% on Humanity’s Last Exam, outperforming Claude 3.7 (8.9%) and DeepSeek R1 (8.6%), but behind OpenAI’s o4-mini (14.3%). It also shines in benchmarks like 78.3% on GPQA and 78.0% on AIME math exams

• The model optimizes reasoning for different tasks—turning off thinking for simple ones and using full reasoning for complex ones. Customers pay for both thinking and output tokens

Try it for F-R-E-E:

• Visit AI Studio and sign in using your Google account

• Once logged in, make sure you're in the Chat section on the left side

• On the right side, select Gemini 2.5 Flash Preview 04-17

• At the middle, enter your query in the text box and press Run

Microsoft

Microsoft Team Develops Revolutionary Low-Resource AI Model for CPUs

Microsoft's BitNet b1.58 2B4T marks a milestone in AI availability, running efficiently on standard CPUs, including Apple's M2. This breakthrough in AI compression could bring advanced AI to everyday hardware.

Details:

• The model quantizes 2 billion parameters to three values (-1, 0, 1), reducing memory use without sacrificing performance. Trained on 4 trillion tokens, it shows efficient design can overcome hardware limits

• BitNet b1.58 2B4T outperformed models like Meta's Llama 3.2 1B, Google's Gemma 3 1B, and Alibaba's Qwen 2.5 1.5B in benchmarks like GSM8K and PIQA, testing math and common sense reasoning

• It is up to twice as fast and uses less memory than similar systems, though it requires Microsoft's custom framework (bitnet.cpp) and lacks full GPU support

Despite GPU compatibility issues, this advancement is a major step toward democratizing AI, enabling powerful capabilities on resource-limited devices without specialized hardware.

Top AI & SaaS Tools

• Jamy (Life-time Deal): Make calls, transcribe them or even video meetings, and turn them into detailed reports so you can focus on the conversation

• Windsurf: This AI coding assistant/IDE is offering unlimited use of GPT-4.1 and o4-mini at no cost from April 14 to 21 [F-R-E-E]

• FIRE-1: Firecrawl’s latest launch that navigates sites, interacts with buttons, fills forms, and finds the exact data you need (Select FIRE-1 from Agent > Model) [F-R-E-E]

• Gamma 2.0: Create impressive AI presentations, interactive websites, social media carousels, and more from simple text prompts [F-R-E-E]

• Omakase: Transform your website into a voice-powered sales agent that talks, listens, and recommends [F-R-E-E]

Start learning AI in 2025

Keeping up with AI is hard – we get it!

That’s why over 1M professionals read Superhuman AI to stay ahead.

• Get daily AI news, tools, and tutorials

• Learn new AI skills you can use at work in 3 mins a day

• Become 10X more productive

Sign up and start mastering AI

Top AI & Tech News

• Google is giving U.S. college students F-R-E-E use of Gemini Advanced, NotebookLM Plus, and 2 TB of storage through spring 2026 finals

• OpenAI's attempt to acquire Cursor coding assistant fell through, so it is in talks to buy Windsurf, another AI coding company, for $3 billion

• Google was found by a federal judge to have illegally monopolized the adtech market by abusing its monopoly power

• Alibaba released Wan 2.1-FLF2V-14B, an open-source model that generates high-quality output from uploaded first and last frame images

• UC Berkeley’s Chatbot Arena is spinning off into a standalone company called LMArena

• OpenAI has launched Flex processing, an API with much lower prices, but slower responses and possible resource unavailability

Interesting Uses of AI

AI Art Spotlight

Model: Midjourney V7

Prompt:

psychedelic album cover. 70s analog sci-fi style but neon iridescent psychedelic futuristic cyberpunk cyborgs --chaos 50 --profile 7noxtdg --stylize 850 --v 7

AI Prompt of the Day

Top AI Video Tutorial

7 AI Skills You MUST Have to Succeed in 2025

Complimentary AI Course of the Day

Building AI Browser Agents

This course covers how autonomous web agents work, their limits, and how to optimize them. You’ll build agents to perform practical tasks like finding, scraping, and summarizing webpages, filling out forms, and subscribing to newsletters—using the AgentQ framework with MCTS, self-critique, and DPO for self-correction.