Alibaba's Qwen3.8-Max, Anthropic Fable Access, and Netflix AI Integration
Alibaba previews its 2.4T parameter Qwen3.8-Max model, Anthropic finalizes Fable 5 access policies, and Netflix reports AI usage across 300 titles. Additionally, OpenAI introduces a new 'useful intelligence' budget scorecard, while SpaceX negotiates a major compute deal with the Pentagon.
Podcast В· 2 min
Alibaba Previews 2.4T Qwen3.8-Max
Alibaba has previewed Qwen3.8-Max, a frontier-scale model featuring 2.4 trillion parameters. The company announced that the full version will be released as open-weight software, allowing developers to download and run the model on their own infrastructure. Alibaba claims the model trails only Claude Fable 5 in performance, though independent benchmarks are not yet available. This release represents a significant move in the open-model space, potentially offering companies an alternative to closed, US-based frontier models.
Anthropic Finalizes Fable 5 Access Policy
Anthropic has ended weeks of uncertainty regarding access to its Fable 5 model. The company confirmed that Fable 5 will remain available to Max and Team Premium subscribers at half of their respective usage caps. Lower-tier plans will receive a one-time $100 credit before transitioning to a pay-per-use model. Anthropic cited unpredictable demand as the reason for the previous access restrictions and stated it is investing in additional compute capacity to stabilize availability.
OpenAI CFO Introduces 'Useful Intelligence' Scorecard
OpenAI CFO Sarah Friar has published a new scorecard for evaluating corporate AI budgets, advocating for a shift from per-token pricing to a metric of 'useful intelligence per dollar.' The framework prioritizes reliability, output quality, and scale over raw token costs. OpenAI illustrated this approach using the DeepSWE coding benchmark, where its Sol model outperformed Fable 5 with a 36.2% lower estimated API bill, highlighting the industry's move toward cost-efficiency and value-based AI adoption.
Netflix Integrates AI into 300 Titles
Netflix has confirmed that generative AI is being used in approximately 300 titles on its platform, primarily for post-production tasks such as creating crowd shots, battle scenes, and worldbuilding elements. Co-CEO Ted Sarandos highlighted that AI-enhanced footage, such as in 'The American Experiment,' can be produced twice as fast and at half the cost of traditional methods. This integration marks a shift toward routine AI usage in media production, potentially enabling smaller teams to create larger-scale content.
SpaceX in Talks for Pentagon Compute Deal
SpaceX is reportedly negotiating a multibillion-dollar deal to provide the U.S. Department of Defense with significant AI data-center capacity. This move aligns with broader trends of major tech and infrastructure firms partnering with government agencies to support AI-driven defense initiatives, further expanding Elon Musk's involvement in government-scale AI infrastructure.
Apple Issues Preservation Letters to Former Employees
Apple has reportedly sent legal preservation letters to approximately 40 former employees who are now working at OpenAI. This action is part of an expanding trade-secrets dispute, with Apple investigating potential unauthorized use of confidential information in the development of AI-related hardware and software.
MLB Bans AI on Dugout iPads
Major League Baseball (MLB) has effectively banned teams from using league-issued dugout iPads for generative AI during games. The decision follows reports of clubs using custom AI applications to assist with in-game strategy, pitch-calling, and substitution decisions. The league's move aims to maintain the integrity of in-game decision-making processes.