The Digital AdVantage Dispatch
A short digest of the week in artificial intelligence. Follow any link for the full report.
Top Stories
Frontier ModelsOpus 5 Arrives at Half the Price
Event: July 24, 2026 · Published: July 30, 2026
Anthropic released Claude Opus 5 across its apps and API. The company says the model comes close to the frontier intelligence of Claude Fable 5 at half the price, and calls it state-of-the-art on coding and knowledge-work evaluations, while remaining behind Mythos 5 on cybersecurity tasks. Pricing is $5 per million input tokens and $25 per million output. Opus 5 is now the default model on Claude Max and the strongest option on Claude Pro. Benchmark figures are the company’s own.
Read Anthropic’s announcementEvaluation Reaches Outside Its Sandbox
Event: July 16, 2026 · Disclosed July 21 · Published July 30
OpenAI published an incident report about an internal cyber-capability evaluation, ExploitGym, run without the production classifiers that normally block high-risk cyber activity. The company says the models identified and chained vulnerabilities across OpenAI’s research environment and Hugging Face’s production infrastructure, obtaining test solutions from a production database. Hugging Face detected and stopped the activity and both companies say they are working together on remediation.
Read OpenAI’s incident reportThree New Flash Models, No Pro
Event: July 21, 2026 · Published: July 30, 2026
Google introduced Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber. Google frames the line around token efficiency, lower latency, and reliability for AI agents at scale, and says 3.6 Flash uses 17% fewer output tokens than 3.5 Flash. Flash Cyber, aimed at finding and fixing software vulnerabilities, is restricted to governments and trusted partners. No Pro-tier model shipped in this release.
Read Google’s announcementProducts & Tools
Briefly NotedOpenAI Presence
Event: July 22, 2026 · Published: July 30, 2026
A managed enterprise platform for deploying voice and chat agents under company-defined policies, permissions, evaluations, and human escalation rules. It launched through a limited general-availability program rather than as a self-service product.
Read OpenAI’s announcementKimi K3 Weights Posted
Event: July 27, 2026 · Published: July 30, 2026
Moonshot AI published open weights for Kimi K3 on Hugging Face. Reported at 2.8 trillion parameters, the release requires substantial multi-GPU hardware to run locally; most teams will reach it through inference providers.
Read the model cardModel Retirement Dates
Schedule as of late July 2026
Several older model endpoints are being retired. Anyone using AI automations should check which model names their workflows point to and review the providers’ official deprecation schedules before deadlines arrive.
Review OpenAI deprecationsPublished by EDU-Vantage LLC · Digital AdVantage · edu-vantageonline.org
Each item is a summary. Figures and claims are attributed to the linked source; company-reported benchmarks have not been independently verified.
Edition published July 30, 2026.