Lead — GPT-5.6’s lower price is a direct bid for production workloads. OpenAI has cut on-demand API pricing for GPT-5.6 Terra by 20% to $2 per million input tokens and $12 per million output tokens, while Luna falls 80% to $0.20 and $1.20 respectively. The change matters because long-running agent systems amplify token costs: a cheaper general-purpose model gives teams more room to keep context, retry tool calls and move routine automation out of prototypes. OpenAI is positioning the reductions as a price-performance move, not a capability change, so buyers should re-run their own workload economics rather than infer a benchmark gain. Read OpenAI’s pricing announcement.
What else matters
ChatGPT Work moves the agent pitch into the product
OpenAI says ChatGPT Work can act across connected apps and files, remain on a project for hours and turn a goal into finished work. It is rolling out on web and mobile to Pro, Enterprise and Edu, with Plus and Business to follow. The practical test is not the demo: organisations will need narrow permissions, audit trails and clear human escalation before handing it consequential workflows. See OpenAI’s ChatGPT Work introduction.
Hugging Face discloses an AI-driven intrusion
Hugging Face says an attacker used malicious dataset code paths to execute code on a processing worker, then escalated access and harvested credentials. It says public models, datasets, Spaces and its published software supply chain showed no evidence of tampering; affected credentials were rotated and the initial code-execution paths closed. The incident turns “agentic attacker” from a tabletop scenario into an operational question for every AI platform that processes untrusted artefacts. Read Hugging Face’s incident disclosure.
Hermes Agent ships a stabilisation release
Nous Research published Hermes Agent v0.19.1, a patch release bundling work merged since v0.19.0 across the gateway, voice, desktop app and installer. The release notes cite roughly 2,789 commits in the window, while deferring the curated change narrative to v0.20.0. For operators, that means treating this as a maintenance update: validate the channels and integrations you actually use rather than expecting a single headline feature. Review the v0.19.1 release notes.
