INFYNY LABSOpenAI releases Terra for everyday computer tasks
OpenAI introduced GPT-6.1 Terra on 26 October. It can interpret screens and take actions through connected tools, as well as draft text and code.
- OpenAI lists form filling and routine work across applications among its uses. It still needs a suitable host, tools and permissions; choosing a model does not connect every app.
- It is available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu, and through the developer API.
- GitHub is also rolling it out gradually in Copilot. Business administrators can control access, so a missing model option may reflect rollout or policy.
- OpenAI still recommends Sol or Astra for the most complex coding and in-depth research. Its performance comparisons are company evaluations.
Try a small task you can check. Our comparison sheet uses a fictional catalogue draft; it contains no completed model tests.
Terra’s lower developer prices need the right comparison
The standard developer price is $1 per million input tokens and $5 per million output tokens. Tokens are pieces of text the AI reads or writes.
- OpenAI says those standard rates are half those of GPT-5.6 Terra and GPT-6.1 Sol. That does not mean a subscription price has halved.
- Cached input, very long inputs, processing options and tools can change charges. How much work an agent does also changes the total.
Compare the complete task, its errors and any human cleanup. A cheaper text rate alone does not establish that a business workflow costs half as much.
Microsoft’s speech service adds more language options
Microsoft announced Azure AI Speech LLM 09-2026 on 26 October. Despite the model name, this is a newly dated announcement about expanded support.
- It adds 25 languages, including 11 Indic languages. Malayalam, Tamil and Marathi are among the listed additions.
- The service turns audio recordings into text and supports multilingual recordings, where speakers switch between supported languages.
- It is available through Fast Transcription and the Foundry playground. Microsoft says the service-side update needs no customer upgrade action.
For a business, the useful test is your actual recording quality, language mix and important names. We have not tested its claimed speed or accuracy gains.
Oracle makes more AI business applications available
Oracle announced 25 new Claw-powered Fusion applications on 26 October, covering work in finance, staffing, supply chains, sales and service.
- Examples include checking payroll differences, reviewing purchasing patterns and preparing follow-up actions.
- The release describes review, permissions and approved actions. It is not an instruction to let software make unrestricted business decisions.
- These are Oracle Fusion offerings, not a free addition to every ChatGPT account. Availability of every adjacent newly announced experience should be checked separately.
Existing Fusion customers can ask which application fits their current process and what setup it requires. No independent savings or deployment result is claimed here.
The UK announces more AI testing and cyber-defence investment
A 26 October ministerial statement sets out additional support for AI testing and cyber security. This is distinct from the Biosecurity Lab covered in our previous edition.
- It includes £120 million more for the AI Security Institute’s testing, research and standards work.
- It also commits £60 million to a National Cyber Shield and £70 million to a centre intended to identify weaknesses across critical infrastructure and the economy.
These are funding commitments and planned capabilities. They do not establish that the protections are operating now or create a new requirement for businesses in India or the GCC.
Anthropic files plans for a first public share offering
Anthropic, the company behind Claude, announced on 26 October that it had filed a US registration statement for a proposed initial public offering.
- It intends to use the Nasdaq symbol ANT. The number of shares and price range are not yet determined.
- The company says the registration statement is not effective and the shares cannot yet be sold through the proposed offering.
This is a filing milestone, not a completed market listing. It does not by itself change access to Claude or establish future product prices.
Extra news: planned nonprofit help and a new Google learning game
Two other 26 October announcements concern people learning and applying AI, rather than a new general-purpose model.
- Taproot and Anthropic plan to match nonprofits with skilled professionals for defined AI projects. The partner announcement describes planned support; it does not establish immediate worldwide eligibility.
- Google launched Polar Peaks, an AI Quests game for ages 11–14. Students practise data cleaning, model testing and human oversight through a brain-mapping story.
- Google says training for teachers covering the third quest follows next month. Earlier teacher modules are separate.
Follow the original pages for access and timing. Neither announcement is evidence that an organisation has already improved its results.