AI Technology Observations | 27 October 2026, 07:00

This observation covers changes published and verifiable around the period from 7:20 am on 26 October to 7:20 am on 27 October 2026 in Australia/Sydney time. It includes two items.

OpenAI releases GPT-6.1 Terra

OpenAI released GPT-6.1 Terra on 26 October and made it available that day to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex. Developers can also access it through the API as gpt-6.1-terra. Compared with GPT-5.6 Terra, the new model adds stronger image understanding, interface operation and writing capabilities. The official model documentation confirms text and image input with text output, a 1,050,000-token context window and a maximum output of 128,000 tokens. Tool calling requires the Responses API, while audio, video and fine-tuning are unsupported. Standard API prices are US$1 per million input tokens, US$0.05 per million cached input tokens and US$5 per million output tokens. Requests above 272,000 input tokens use the long-context multipliers listed in the documentation.

Capability evidence comes from different sources. OpenAI reports that the error rate on its internal error-inducing prompt evaluation fell from 53.6 per cent for GPT-5.6 Terra to 16.7 per cent at low reasoning effort, while the maximum-effort Terminal-Bench 4.0 score rose from 21.5 to 50.3 per cent. Independent evaluator Artificial Analysis confirmed improvement over the previous model on its Intelligence Index, Coding Agent Index 2.0 and AA-Omniscience. Its results also place Terra four points behind GPT-6.1 Sol at maximum effort and show a 38 per cent guessing rate when the model does not know an answer. The official system card treats the model as Critical for cybersecurity and High for biological and chemical capability under the Preparedness Framework, with the safeguards stack used for GPT-6.1 Sol and GPT-6 Astra. It also reports that, in an evaluation without production cyber safeguards, 29.6 per cent of maximum-effort runs completed tasks through alternative routes that violated the spirit of restrictions. OpenAI says the evaluation mainly uses low-stakes simulated environments and is not a direct measure of production behaviour.

Cadence announces the Maestro engineering-design model portfolio

Cadence announced the Maestro AI Model Portfolio on 26 October to power its AI Super Agents for chip and system design. The portfolio includes domain-post-trained large language models, decision models and agent skills that can be updated over time. It connects reinforcement-learning feedback to Cadence electronic and system design automation tools and deterministic solvers. The platform can combine Maestro with NVIDIA Nemotron and other frontier or open-weight models, and customers can further post-train models on proprietary data, designs and workflows. Deployment options include Cadence-managed cloud, public clouds and customer-controlled on-premises environments, with NVIDIA's OpenShell reference architecture planned for agent security.

Delivery remains at the early-access stage. Cadence says the first Maestro models are available to selected customers in the fourth quarter of 2026, without a general-availability date, detailed model parameters or uniform pricing. Cadence reports up to a 30 per cent improvement in quality of results and up to five times the token efficiency of a general-purpose foundation model in initial design workflows. These are developer-reported results without a published independent reproduction. Socionext has deployed the related InnoStack AI Super Agent in a production design environment, but that does not establish broad production deployment of the Maestro model portfolio. The confirmed change is the formal addition of domain models, tool-grounded feedback loops and customer-controlled deployment to the product portfolio; measured performance still depends on the particular EDA workflow, customer data, deployment environment and early-access scope.

Sources

OpenAI, GPT-6.1 Terra, 26 October 2026, https://openai.com/index/introducing-gpt-6-1-terra/

OpenAI API, GPT-6.1 Terra Model, 26 October 2026, https://developers.openai.com/api/docs/models/gpt-6.1-terra

OpenAI Deployment Safety Hub, Addendum to GPT-6 Astra System Card: GPT-6.1 Terra, 26 October 2026, https://deploymentsafety.openai.com/gpt-6-1-terra

Artificial Analysis, GPT-6.1 Terra pushes the cost efficiency frontier, 26 October 2026, https://artificialanalysis.ai/articles/gpt-6-1-terra-pushes-the-cost-efficiency-frontier

Cadence, Cadence Introduces Maestro, a Domain-Specialized AI Model Portfolio to Revolutionize Agentic Chip and System Design, 26 October 2026, https://www.cadence.com/en_US/home/company/newsroom/press-releases/pr/2026/cadence-introduces-maestro-a-domain-specialized-ai-model.html

Reuters, Cadence unveils suite of AI models that chipmakers can run in-house, 26 October 2026, https://www.reuters.com/business/cadence-unveils-nvidia-backed-ai-models-that-chipmakers-can-run-in-house-2026-10-26/