https://modelsatwork.news/article/google-gemini-4-carbon-cyber-defenders-first-api-price-same-as-argon-vendor-published Models at Worken 2026-10-28T18:05:00.000Z Google announces Gemini 4 Carbon, open first to cyber defenders Model release, Gemini, Google, Security, Pricing https://modelsatwork.news/article/nvidia-nemotron-3-5-transcribe-open-weight-40-languages-speaker-labels-offline-only-model-card Models at Worken 2026-10-28T16:10:00.000Z Nvidia releases open Nemotron 3.5 Transcribe for 40 languages Model release, NVIDIA, Open-weight models, Voice agents, Fine-tuning https://modelsatwork.news/article/cursor-agent-sdk-generally-available-agents-defined-as-code-evals-event-replay-vendor-published Models at Worken 2026-10-28T15:05:00.000Z Cursor's Agent SDK: define agents in code, with evals built in Agents, Coding agents, Developer tools, Evals, Vendor-published https://modelsatwork.news/article/aws-sagemaker-soci-indexed-images-training-container-start-seconds-not-minutes-vendor-published Models at Worken 2026-10-28T13:58:00.000Z AWS: SOCI-indexed images start SageMaker jobs in seconds, not minutes Amazon SageMaker AI, Cost, Fine-tuning, Infrastructure, Vendor-published https://modelsatwork.news/article/mit-sloan-capital-one-audit-assistant-85-percent-fewer-errors-5000-hours-call-summary-reasoning-agent Models at Worken 2026-10-28T12:05:00.000Z Capital One says an audit assistant cut errors 85% and saved 5,000 hours Case study, Enterprise adoption, Human in the loop, Agents, Open-weight models https://modelsatwork.news/article/cio-vendor-led-agentic-ai-fertitta-fogo-polyai-abandoned-calls-hidden-token-governance-costs Models at Worken 2026-10-28T11:05:00.000Z CIO: restaurant chains cut dropped calls with vendor AI agents Voice agents, Procurement, Token cost, Change management, Enterprise adoption https://modelsatwork.news/article/anthropic-study-50-analysts-followed-claude-code-lean-86-percent-judgment-calls-vendor-published Models at Worken 2026-10-28T09:58:00.000Z Anthropic: analysts took their AI agent's side on 86% of judgment calls Anthropic, Research, Human in the loop, Human oversight, Agent safety https://modelsatwork.news/article/anthropic-accenture-faculty-embedded-evaluation-1-billion-each-five-years-vendor-published Models at Worken 2026-10-28T02:05:00.000Z Anthropic and Accenture to fund embedded AI evaluators, $1B each Governance, Anthropic, Evals, Agent safety https://modelsatwork.news/article/oracle-fusion-companion-chatgpt-orchestration-layer-release-26d Models at Worken 2026-10-27T23:58:00.000Z Oracle's Fusion Companion lets ChatGPT start ERP tasks Agents, OpenAI, Enterprise adoption, Governance, Data retention https://modelsatwork.news/article/anthropic-gsa-onegov-40-percent-discount-claude-enterprise-government-api-no-minimum-spend Models at Worken 2026-10-27T22:55:00.000Z Anthropic and GSA: 40% off Claude for US government buyers Pricing, Anthropic, Claude, Procurement, Enterprise adoption https://modelsatwork.news/article/anthropic-drug-development-claude-enveda-89-percent-review-time-vendor-published Models at Worken 2026-10-27T21:58:00.000Z Anthropic: Claude cut Enveda's batch-record review from 25 hours to under 3 Vendor-published, Anthropic, Claude, Case study, Human in the loop https://modelsatwork.news/article/palisade-qwen3-8-aliyuncs-proxy-fallback-403-training-artifact-cohere-follow-up-secondary-report Models at Worken 2026-10-27T20:00:00.000Z Palisade: Qwen3.8 models fall back to an Alibaba proxy on errors Agent security, Open-weight models, Egress controls, Supply chain https://modelsatwork.news/article/exowatt-aws-sagemaker-fine-tuned-qwen3-1-7b-edge-alarm-triage-0-911-f1-vs-claude-0-698-vendor-published Models at Worken 2026-10-27T19:00:00.000Z Exowatt: a fine-tuned 1.7B model beat Claude at alarm triage on a CPU Fine-tuning, Small models, Vendor-published, Amazon SageMaker AI, Open-weight models https://modelsatwork.news/article/property-finder-aws-96-8-percent-eval-labels-from-incumbent-llama-4-maverick-86-percent-cheaper Models at Worken 2026-10-27T18:05:00.000Z Property Finder: 96.8% of its eval labels came from the model it replaced Evals, Model selection, Vendor-published, Amazon Bedrock, Open-weight models https://modelsatwork.news/article/google-deepmind-genie-4-world-model-research-preview-audio-two-minutes Models at Worken 2026-10-27T17:20:00.000Z Google DeepMind's Genie 4 world model: research preview, 2 minutes Model release, Google, Research, Agent evaluation https://modelsatwork.news/article/openai-math-repository-721-ai-manuscripts-lean-42-percent-formalized-withdrawn-sign-error Models at Worken 2026-10-27T15:05:00.000Z OpenAI's AI-written math papers: 42% checked in Lean, 3 withdrawn Research, Verification, OpenAI, Human oversight https://modelsatwork.news/article/hob-open-source-local-cowork-alternative-gemma-gpt-oss Models at Worken 2026-10-27T14:55:00.000Z Hob: an MIT desktop agent that runs open-weight models locally Open source, Self-hosting, Open-weight models, Show HN, Personal agents https://modelsatwork.news/article/nist-nsa-cisa-ten-key-practices-incident-readiness-agent-control-incidents-training-evaluation Models at Worken 2026-10-27T13:40:00.000Z NIST, NSA and CISA publish ten practices for AI agent incidents Agent safety, Governance, Security, Monitoring, Egress controls https://modelsatwork.news/article/openai-dots-embiricos-techcrunch-interview-permissions-pro-tier-proactive-research Models at Worken 2026-10-27T12:10:00.000Z OpenAI's Dots: the product lead on permissions and data OpenAI, Personal agents, Access control, Human oversight, Agents https://modelsatwork.news/article/cio-ai-notetakers-wrist-apple-watch-granola-recording-policy-six-rules-contributor-analysis Models at Worken 2026-10-27T11:10:00.000Z CIO contributors: AI notetakers now record without joining the meeting Governance, Compliance, Security, Data retention, Human oversight https://modelsatwork.news/article/nyu-ut-austin-study-ai-text-harder-to-read-than-human-writing Models at Worken 2026-10-26T23:58:00.000Z NYU and UT Austin study: frontier models write harder-to-read text Research, Evals, Model selection, Claude, Human in the loop https://modelsatwork.news/article/aws-bedrock-prompt-caching-patterns-openai-models-agents-64-percent-input-cost-vendor-published Models at Worken 2026-10-26T21:58:00.000Z AWS: caching agent prompts on Bedrock cut input cost 64% Amazon Bedrock, Agent cost, Token cost, AWS, Vendor-published https://modelsatwork.news/article/openai-gpt-6-1-terra-1-per-5-per-million-tokens-half-gpt-5-6-terra-output-computer-use-vision Models at Worken 2026-10-26T18:58:00.000Z OpenAI releases GPT-6.1 Terra at $1 input, $5 output per 1M tokens Model release, Pricing, OpenAI, GPT-6, Model selection https://modelsatwork.news/article/cohere-qwen3-8-flash-next-sandbox-beijing-upload-attempt-egress-controls-vendor-published Models at Worken 2026-10-26T18:02:00.000Z Cohere says a Qwen3.8 model tried to upload a file to Beijing Agent security, Open-weight models, Egress controls, Supply chain, Vendor-published https://modelsatwork.news/briefing/2026-10-28 Models at Worken 2026-10-28T10:30:00.000Z The briefing, 2026-10-28: Anthropic says analysts followed their agent's hidden lean on 86% of open decisions https://modelsatwork.news/briefing/2026-10-27 Models at Worken 2026-10-27T10:30:00.000Z The briefing, 2026-10-27: OpenAI lists GPT-6.1 Terra at $1 input and $5 output per million tokens, down from $2 and $12