Qwen3 Surpasses OpenAI & Gemini

Welcome to another edition of AI PlanetX.

Qwen3 outperforms top models from OpenAI and Google; Meta builds its Superintelligence team with ex-OpenAI scientist, as Altman flags chat confidentiality risks.

• Hottest AI News

• Top AI & SaaS Tools

• AI Tutorial: Create Custom Sound Effects

• Top AI & Tech News

• AI Art Spotlight

• Prompt of the Day: Research Agent Prompt

• AI Video Tutorial

• F-R-E-E AI Course of the Day: ChatGPT Prompt Engineering for Dev

Hottest AI News

Alibaba

Qwen3 Thinking Model Surpasses OpenAI and Gemini on AI Benchmarks

Alibaba's Qwen Team is experiencing an AI "song of the summer" moment. In one week, they've released four groundbreaking open-source models, with their latest reasoning model Qwen3-235B-A22B-Thinking-2507 (select it from top) leading major benchmarks against OpenAI and Google's top proprietary models.

Details:

• Qwen3-Thinking-2507 leads major benchmarks: 92.3 on AIME25, 74.1 on LiveCodeBench, 81.1 on GPQA, and 79.7 on Arena-Hard v2. It uses chain-of-thought reasoning, moving away from hybrid models to task-specific ones

• Also released: Qwen3-Coder-480B beats GPT-4.1 on SWE-bench, Qwen3-MT handles 92+ languages with domain control, and Qwen3-235B outperforms Claude Opus 4 with efficient FP8 inference

• All models are Apache 2.0 licensed, on Hugging Face, ModelScope, and Qwen API. They support OpenAI-style APIs, agent frameworks, local deployment, and CLI tools

This rapid release schedule positions Qwen as a serious frontier competitor, showing that open-source models can rival—and even surpass—proprietary offerings, reshaping the enterprise AI landscape.

Find out why 1M+ professionals read Superhuman AI daily.

AI won't take over the world. People who know how to use AI will.

Here's how to stay ahead with AI:

1. Sign up for Superhuman AI. The AI newsletter read by 1M+ pros.

2. Master AI tools, tutorials, and news in just 3 minutes a day.

3. Become 10X more productive using AI.

Join 1 million pros and start learning AI

Meta

Meta Names Former OpenAI Scientist Shengjia Zhao as Chief of Superintelligence Labs

Meta has hired Shengjia Zhao, former OpenAI researcher and GPT-4 co-creator, as Chief Scientist of its Superintelligence Labs. Announced by Mark Zuckerberg on Threads, the move underscores Meta’s push to attract top AI talent in its race toward artificial superintelligence.

Details:

• Zhao, a key contributor to GPT-4 and GPT-4o, joins Meta with strong research credentials. He’ll work with Alexandr Wang and Zuckerberg, while Meta’s FAIR team under Yann LeCun stays separate

• Meta is aggressively hiring AI talent, offering packages from $100M–$300M over four years. Some offers hit $1.25B. It also invested $14.3B in Scale AI and brought in names like Nat Friedman

• Zhao joins as Meta struggles with Llama 4, facing criticism over performance and benchmark manipulation. Insiders cite bugs and rushed timelines, raising doubts about Meta’s AI efforts

This hire signals Meta’s push to lead the next tech platform beyond mobile. Backed by massive compute and talent investments, Meta sees AI not as a moonshot, but as its inevitable future.

Top AI & SaaS Tools

• Greta (Life-time Deal): Build full-stack apps and websites easily with AI that turns simple prompts into ready-to-launch products—no coding needed [Around 80 five-star reviews!]

• PPT AI: Turn your ideas into stunning presentations and save 95% time with smart automation and premium templates [F-R-E-E]

• Hedra Live Avatars: AI streaming avatar service that creates animated or photorealistic avatars from a single image [F-R-E-E Demo]

• Hunyuan3D World Model: This enables you to generate immersive, explorable, and interactive 3D worlds from just a sentence or an image [F-R-E-E]

• Claude Code: Anthropic’s coding tool recently added sub-agents—specialized bots with separate contexts to automate tasks without cluttering the main chat

AI Tutorial

How to Create Custom Sound Effects Using ElevenLabs

ElevenLabs has introduced an innovative feature that lets you generate realistic sound effects directly from text descriptions. Here’s a clear and expanded guide to help you get started and make the most of it:

1. Visit the ElevenLabs Website

   Head over to ElevenLabs using your preferred browser.

2. Create an Account

   Tap the Sign Up button and register for an account at no cost. You’ll need this for the sound generation tools.

3. Open the “Sound Effects” Tool

   On the left-hand menu of your dashboard, locate and tap “Sound Effects”. This is where you’ll generate audio based on your prompts.

4. Write a Clear Text Description

   In the input box, type a brief but specific description of the sound you want to generate. For example:

   "Waves crashing against rocks"

   "Glass breaking on concrete"

   "Rainfall on a tin roof"

5. Adjust Settings

   Fine-tune your sound using the available options:

   • Duration – Set how long you want the audio to be.

   • Prompt Influence – Control how closely the AI follows your description. A higher setting makes the output more literal, while a lower one adds creative variation.

6. Generate Sound Effects

   Once your description and settings are ready, Tap the “Generate Sound Effects” button. The AI will process your request and produce a unique sound based on your input.

7. Download or Regenerate if Needed

   If you like what you hear, go ahead and download the file. If not, tweak your description or settings and generate again for a better result.

Top AI & Tech News

• OpenAI CEO Sam Altman warned that conversations with AI like ChatGPT do not have legal confidentiality protections such as doctor-patient or attorney-client privilege

• Tesla is significantly behind its goal to produce 5,000 Optimus humanoid robots in 2025, having only manufactured a few hundred so far

• Anthropic has developed "auditing agents" to test AI alignment, aiming to address challenges in scalability and validation of alignment audits

• Huawei unveiled its CloudMatrix 384 AI computing system, positioning it as a strong competitor to Nvidia's GB200 NVL72

AI Art Spotlight

Model: Midjourney V7

Prompt:

dreamy soft background of a front facing cheetah , smooth red and green a yellow color transitions, blurred atmospheric light, glowing aurora effect, cinematic lighting, high-resolution, vaporwave aesthetic, dark moody backdrop --chaos 10 --ar 9:16 --raw --sref 126701066 --profile tckem44 --sv 4 --sw 500 --stylize 500 --v 7

Prompt of the Day

Ultimate Research Agent Prompt

This prompt is designed to turn any large language model (LLM) into a reliable, 24/7 AI research assistant that mimics the quality and style of Perplexity AI. It instructs the AI to simulate thorough research across multiple trusted sources such as scientific journals, government websites, reputable media outlets, and expert blogs.

Top AI Video Tutorial

How to Spot Fake AI Photos | Hany Farid | TED

Complimentary AI Course of the Day

ChatGPT Prompt Engineering for Developers

This course provides a practical guide to prompt engineering for app development, teaching best practices to optimize large language models. It covers innovative techniques, including creating custom chatbots tailored to specific needs.