Happy Wednesday. Google's Gemini 4 Carbon landed Monday at a quarter of Anthropic's Claude Fable 5.2 price, with a December 1 deadline before rates double. That pricing decision is real and time-sensitive for any team running cost-sensitive AI pipelines.
Elsewhere, PayPal's AI team offers a sharp lesson in where enterprise AI value actually lives, and OpenAI is turning ChatGPT into a storefront, with 1.2 billion weekly users as the sales pitch.
ARTIFICIAL INTELLIGENCE
Google's Carbon Model Undercuts Claude by 75%, but Only Until Dec. 1
Google opened Gemini 4 Carbon to all paying developers on Nov. 10 at $2 per million input tokens and $10 per million output tokens. That is a quarter of what Anthropic charges for Claude Fable 5.2, which launched a day earlier at $8 input and $40 output. On Dec. 1, Carbon's rates double to $4 and $20.
The benchmark picture complicates the buy decision. Fable 5.2 leads Carbon in six of seven head-to-head categories, and Claude Code with Opus 5.5 scores 62 on the Coding Agent Index against Carbon's 59. The price gap is real and the introductory window is short. Teams with high token volume and cost pressure have about three weeks to decide whether Carbon's capability is good enough at half the long-term price, or whether the benchmark gap with Fable 5.2 justifies paying more after December.
ARTIFICIAL INTELLIGENCE
PayPal Reached 95% AI Accuracy Without Touching the Model
PayPal's internal data science assistant jumped from 43% to 95% accuracy by adding golden queries and analytical playbooks stored outside the LLM, not by fine-tuning or switching models. The assistant logged more than 1,000 questions and saved an estimated 1,600 hours over 10 weeks. The underlying Cosmos gateway now processes 200 billion tokens per day across more than 100 models at 99.999% availability.
The architecture choice has a direct cost implication for any enterprise AI team: keeping institutional knowledge in a separate layer means a new model release does not require rebuilding that knowledge from scratch. PayPal's experience also shows the limits of fine-tuning smaller models for latency, where an 8-billion-parameter model cut lag but hurt accuracy even after task-specific fine-tuning. The governance note is worth filing: in a system processing $1.8 trillion annually, PayPal still keeps humans in the loop on agentic workflows.
ARTIFICIAL INTELLIGENCE
OpenAI Opens ChatGPT to Business Profiles and AI Sales Agents
OpenAI launched Business Hub and Business Agents in beta on Nov. 10, letting companies claim profiles inside ChatGPT, publish AI sales tools, and see aggregated mention analytics. Both products are free at launch. Shopify merchants get a direct integration path, with eligible products already pulled into ChatGPT through the Shopify Catalog. The program is currently limited to U.S. e-commerce and online service businesses.
OpenAI says claiming a profile does not influence how often a business appears in ChatGPT answers, a distinction that matters if you are thinking about this as a search optimization play. It is not. The agent activates only when a user clicks a business profile, which is a narrower surface than it sounds. The policy language is also worth reading before building on this: OpenAI does not verify every business-supplied detail, does not guarantee agent accuracy, and can include third-party information in profiles. The free tier and no announced paid structure make the monetization path unclear.
Created by the robots at The SaaS Sentinel