🎙️ AI 日報 Podcast，Apple起訴OpenAI、GPT-5.6與Gemini 3.5大戰、OpenAI估值破兆IPO

Opening Remarks

Hello everyone, good morning! I'm Mark. Today's topics can only be described as a "clash of the titans"—the tech world is absolutely exploding. From Apple suing OpenAI for poaching 400 employees and stealing trade secrets, to the simultaneous head-to-head showdown among major model giants with GPT-5.6, Gemini 3.5 Pro, Grok 4.5, and Claude Fable 5, and finally, OpenAI's IPO plans with a valuation eyeing one trillion dollars. We have a massive amount of information today, so get your coffee ready, and let's dive right in!

Today's Top Stories

1. Tech Giants Face Off! Apple Sues OpenAI for Stealing Trade Secrets

Source: BuildFastWithAI (https://www.buildfastwithai.com/blogs/ai-news-today-july-12-2026）
Summary: Apple has officially filed a lawsuit against OpenAI in California federal court, accusing the company of systematically and deliberately poaching over 400 Apple employees. The scope of the poaching covers chip design, edge AI, and hardware development. Apple contends that this is not normal talent mobility, but rather a carefully orchestrated theft of technology.
Taiwan Perspective: This lawsuit has a massive impact on Taiwan's semiconductor and IC design industries. Apple's chips and edge AI technology have always been tightly integrated. If OpenAI indeed uses this to make up for its hardware shortcomings, Taiwanese manufacturers may have to reshuffle their positions in the supply chains of these two major camps in the future.
Discussion Points:
- Where is the line for "reasonable poaching" when it comes to a cross-company movement of 400 employees?
- Is Apple, which is actively positioning itself in "on-device AI," using its legal iron fist because it feels threatened by OpenAI?
Script Suggestion: This is truly spectacular! Apple is taking off the gloves this time and taking OpenAI straight to court. Think about it—400 top engineers. That is practically moving Apple's entire elite chip and on-device AI team lock, stock, and barrel. Everyone knows Apple's A-series and M-series chips are powerful precisely because of their on-device computing, whereas OpenAI has historically lacked hardware DNA. No matter who wins or loses this lawsuit, it signifies that the Silicon Valley talent war has reached a fever pitch. For us engineers in Taiwan's Hsinchu Science Park, this is also an excellent opportunity to observe the future direction of edge computing.

2. Censorship Lifted! OpenAI's GPT-5.6 Family Fully Launched

Source: CNBC (https://www.cnbc.com/2026/07/08/openai-expanding-gpt-5point6-ai-model-release-ending-government-limits.html）
Summary: After cooperating with the US government for two weeks of safety reviews and red-teaming, OpenAI has finally announced the full release of its GPT-5.6 model family (Luna, Terra, Sol). It is now live on ChatGPT, Codex, and the API. The top-tier Sol model is priced at $5 p e r m i l l i o n i n p u t t o k e n s a n d$ 30 per million output tokens.
Taiwan Perspective: The pricing of the top-tier Sol model represents a significant cost burden for Taiwan's small and medium-sized startups and indie developers. Everyone will need to calculate token efficiency more precisely when calling the API.
Discussion Points:
- Government intervention in AI model reviews is becoming the norm. Will this slow down or accelerate technological innovation?
- How do the three differently positioned models (Luna, Terra, Sol) help developers in terms of choice?
Script Suggestion: After much anticipation, GPT-5.6 has finally cleared the US government's hurdle and is fully open! Among the three musketeers released this time, the strongest one, Sol, has incredible computing power, but its price is definitely not cheap. When integrating the API, our developer community in Taiwan might want to carefully evaluate whether lightweight tasks can just use Luna or Terra. Spend your budget where it matters most; otherwise, the accumulation rate per million tokens will quickly drain a startup's wallet.

3. A Massive 2-Million Token Appetite! Google's Gemini 3.5 Pro Set to Debut

Source: BuildFastWithAI (https://www.buildfastwithai.com/blogs/ai-news-today-july-12-2026）
Summary: Google DeepMind has announced that it will launch Gemini 3.5 Pro on July 17. The biggest highlight is its massive 2-million token context window, directly doubling the current industry standard. Furthermore, its "Deep Think" mode, which focuses on deep reasoning, will be offered exclusively to Ultra subscribers at $250 per month.
Taiwan Perspective: For Taiwanese enterprises that need to process long Traditional Chinese legal regulations, medical records, or large-scale project source code, a 2-million token capacity is practically a godsend. It saves a lot of trouble with chunking and vector databases.
Discussion Points:
- Will a 2-million token context length completely replace traditional RAG (Retrieval-Augmented Generation) technology?
- Is the $250 monthly subscription fee cost-effective enough for enterprise users and individual developers?
Script Suggestion: Google is really gunning for OpenAI this time! Gemini 3.5 Pro stretches the context window to 2 million tokens in one go. This means you can throw an entire year's worth of corporate financial reports, or dozens of thick Chinese user manuals, in all at once, and it won't forget a thing. However, that $250-a-month "Deep Think" feature translates to nearly 8,000 NTD. This pricing strategy is clearly aimed at harvesting enterprise users; individual users will probably just watch from the sidelines.

4. The Disrupter Arrives! xAI Releases Grok 4.5, Featuring High Cost-Performance and Real-Time Data

Source: LLM Stats (https://llm-stats.com/llm-updates）
Summary: Elon Musk's xAI has launched Grok 4.5. It not only adds deep reasoning capabilities comparable to Anthropic's Opus level but also retains its signature real-time X platform data integration. Most importantly, Grok 4.5 maintains relaxed safety guardrails, earning it the title of the most cost-effective flagship model on the planet right now.
Taiwan Perspective: For social media marketers or public opinion analysts in Taiwan, Grok 4.5's real-time social data analysis capability is an indispensable tool for tracking international current events and social media trends.
Discussion Points:
- While permissive guardrails are convenient, will they bring greater information security and fake news risks?
- Will Musk's xAI low-price market penetration strategy disrupt the existing LLM market pricing order?
Script Suggestion: Elon Musk's Grok 4.5 is truly playing the role of a disrupter this time. Its reasoning capabilities are catching up fast with the top tier, and its best trick is still the ability to pull real-time data from the X platform. Plus, unlike GPT or Claude, it doesn't come with a heavy sense of moral policing and restrictions, making it much less constrained to use. The key is, its price is incredibly competitive! For Taiwanese social media editors or startup teams who want to do real-time public opinion analysis without spending a fortune, Grok 4.5 is definitely worth a try.

5. Valuation Heading for $1 Trillion! OpenAI Secretly Files for IPO

Source: CNBC (https://www.cnbc.com/2026/05/20/openai-ipo-filing.html）
Summary: Sources indicate that OpenAI, through Goldman Sachs and Morgan Stanley, has secretly filed its registration statement (S-1) for an initial public offering with the US SEC, with plans to list as early as the fourth quarter of 2026. Currently, OpenAI's monthly revenue has reached $2 b i l l i o n . A n a l y s t s e s t i m a t e i t s p o s t - I P O m a r k e t v a l u e w i l l s u r p a s s$ 850 billion and could even eye $1 trillion, making it one of the top five largest IPOs in US history.
Taiwan Perspective: OpenAI's IPO will be the ultimate litmus test for the global AI bubble debate. As Taiwan is a major hub for hardware and server manufacturing, Taiwan's AI concept stocks will inevitably experience wild fluctuations alongside this IPO of the century.
Discussion Points:
- Is $2 b i l l i o n i n m o n t h l y r e v e n u e e n o u g h t o s u p p o r t i t s m a s s i v e c o m p u t i n g c o s t s a n d a$ 1 trillion valuation?
- What does OpenAI's transition from a "non-profit organization" to a public listing mean for the open-source community and the democratization of AI?
Script Suggestion: This is absolutely the most shocking financial news of the year, if not the decade! OpenAI is actually going public, and with a valuation looking at one trillion dollars. What does that even mean? That is equivalent to the scale of a TSMC! Although they can make $2 billion a month, don't forget that the money they burn on computing power is also astronomical. Once OpenAI's IPO succeeds, it will completely cement AI's position as the next generation of infrastructure. Taiwan's major hardware manufacturers making servers and cooling systems are probably going to have endless orders to fill.

Closing Paragraph

Alright, did these five heavyweight stories get everyone excited and perhaps a little anxious? We saw the patent and talent war between Apple and OpenAI, witnessed the close-quarters combat among the four major models in terms of performance, price, and context length, and observed AI commercialization about to reach a historic one-trillion-dollar milestone. The future AI world will only change faster, so make sure to stay tuned to our show. See you next time, bye-bye!