🤖 Claude Fable 5 stays at 50% for premium plans 🤖
Anthropic finally settles Fable 5 access after three delays in five weeks.

After three postponements in just five weeks, Anthropic confirmed that Claude Fable 5 will remain available inside Max and Team Premium plans at 50% of each plan's usage caps. Lower-tier subscribers on Pro and Team Standard will receive a one-time $100 credit before transitioning to pay-per-use pricing. Anthropic acknowledged that demand has been "challenging to predict" and called the access situation "frustrating" — a rare public admission. OpenAI CEO Sam Altman publicly mocked Anthropic's handling of Fable 5 access on X, adding salt to the wound. Competitive pressure mounts as rivals expand access limits: Moonshot's Kimi K3 is approaching frontier performance, and OpenAI continues broadening usage for GPT-5.6 Sol. Anthropic says it is investing in additional compute to improve availability as the rollout continues.

🎙️ Google brings Gemini Live and Skills to web 🎙️
Google's conversational voice AI expands beyond mobile with a full web rollout.

Google is preparing to bring Gemini Live — its real-time conversational voice AI — to desktop and web browsers, extending the feature beyond its current mobile-only availability. The rollout would make Gemini Live accessible across all major platforms, significantly broadening its user base and increasing competitive pressure on OpenAI's Advanced Voice and Anthropic's voice-enabled Claude. The addition of Skills functionality suggests deeper agentic and task-completion capabilities are coming to the web experience. Cross-platform availability is fast becoming a key differentiator in conversational AI, and Google is clearly intent on defending its distribution advantage across devices.

💰 OpenAI CFO pitches 'Intelligence Per Dollar' AI metric 💰
OpenAI's CFO reframes enterprise AI pricing around output value, not token cost.

OpenAI CFO Sarah Friar published a framework proposing "useful intelligence per dollar" as a new standard to replace per-token pricing comparisons. The scorecard measures reliability, useful output quality, scale, and total expense. To demonstrate, OpenAI used the DeepSWE coding benchmark, where GPT-5.6 Sol scored 2.8 points above Anthropic's Fable 5 while estimating a 36.2% lower API cost. Friar's central question: whether "the value of the work AI completes grows faster than the cost of producing it." Cost efficiency now matters as much as raw capability in the enterprise market — and this framework is a clear strategic positioning move by OpenAI to redefine how buyers compare providers.

INTERESTING TO KNOW

⚖️ Apple targets 400+ OpenAI staff in trade secret suit ⚖️

As regulation of talent mobility faces new scrutiny, Apple has sent legal preservation letters to dozens of former employees now working at OpenAI, instructing them to preserve documents relevant to a trade secret lawsuit. Apple alleges OpenAI systematically recruited engineers who brought proprietary knowledge of designs and manufacturing processes. More than 400 former Apple employees currently work at OpenAI, underscoring the massive scale of talent migration. The case could reshape employee mobility and IP protections across the industry.

🚀 Alibaba ships Qwen3.8 with 2.4 trillion parameters 🚀

In a rollout that could reshape the open-weight landscape, Alibaba unveiled Qwen3.8, a 2.4-trillion-parameter open-weight model claiming near-frontier performance "second only to Claude Fable 5." A preview is already accessible through Alibaba's Token Plan, Qoder, and QoderWork platforms. At this scale, developers and enterprises gain access to near-frontier capabilities without proprietary API lock-in — a direct competitive challenge to Western closed-source providers.