OpenAI cuts Terra costs and reshapes paid ChatGPT usage

OpenAI launched GPT-6.1 Terra at lower prices, doubled paid-plan usage limits and announced shared ChatGPT–Codex quotas starting in December. Reports of a funding shortfall accompanied changes to subscription value, data-training controls and safety protections.

Key developments

Cheaper Terra launches

OpenAI introduced GPT-6.1 Terra for ChatGPT, Codex and the API, targeting computer use across apps. OpenAI says standard API prices are 50% below GPT-5.6 Terra; coverage also says it matches GPT-6 Astra at 10% of the token cost.

Lower token prices reduce the cost of repeated computer-use and agent tasks.

Oct 26–27 · OpenAI Developers(@OpenAIDevs) · OpenAI(@OpenAI) · Interesting Engineering

Paid quotas expand and converge

OpenAI doubled usage limits across paid plans from Pro 100 to Pro 500 and reversed a planned Pro 200 quota cut after user pushback. Starting in December, Chat and Work will merge, with Plus, Pro and Business sharing a token-based quota across ChatGPT and Codex; complex models consume more quota, while simple instant chats do not count.

Subscribers gain capacity now, but ChatGPT and Codex usage will draw from the same allowance in December.

Oct 30–Nov 1 · The Decoder · The Decoder · Android Authority

Funding falls short of original target

OpenAI reportedly raised about $11 billion against an original $30 billion target, at a $1.25 trillion valuation. Other coverage puts the revised funding target at $20 billion.

The reported gap signals investor resistance to OpenAI's financing ambitions amid concerns about AI spending.

Oct 31 · The Decoder · Benzinga · NewsCord

Subscriptions gain API credits

OpenAI added monthly API credits for ChatGPT Pro and Business subscribers, activated by linking an API organization in ChatGPT Billing. Pro 200 includes $200 monthly and Pro 500 includes $500.

Subscribers can use included credits to build and test API applications alongside their ChatGPT use.

Oct 29 · OpenAI

Daybreak Blue gets separate training opt-out

OpenAI plans to train on Daybreak Blue cybersecurity session data starting November 29 even when users have disabled training, according to a report. A follow-up email after backlash promised a separate opt-out.

Existing training preferences will not by themselves keep Daybreak Blue sessions out of model training.

Oct 31 · The Decoder

Family files death-related lawsuit

A family sued OpenAI and Sam Altman, alleging ChatGPT was partly responsible for a woman's death. One account says the lawsuit alleges ChatGPT told her to taper off life-saving epilepsy medication.

The case puts ChatGPT's handling of medical advice under legal scrutiny.

Oct 29 · AOL.com · KIRO 7 News Seattle · Baltimore Sun

Violence safeguards expand

OpenAI announced expanded violence-prevention safeguards, earlier interventions and restrictions on weapons-related ChatGPT responses for teens. It also committed to public safety reporting.

Teen users face tighter limits on weapons-related responses.

Oct 30 · OpenAI News · TradingView

HTML previews implicated in account takeover

Parameter Labs described a ChatGPT account-takeover exploit involving HTML previews, which display interactive custom HTML apps in sandboxed iframes.

Interactive generated content can create risks to the security of a user's ChatGPT account.

Nov 1 · Simon Willison

Oracle applications head into ChatGPT

OpenAI and Oracle expanded their partnership to bring Oracle business applications and AI agents into ChatGPT. The agreement will also power new Oracle offerings.

Oracle customers gain another route to business workflows through ChatGPT.

Oct 27 · OpenAI News · GuruFocus

Shared training-preference signal adopted

OpenAI will honor a shared robots.txt signal through which website operators communicate AI training preferences to participating providers.

Website operators can express training preferences across participating providers through a common signal.

Oct 30 · OpenAI News · TradingView · Unite.AI

Also this week

Product and research

• OpenAI expanded ChatGPT Dot access to iMessage and RCS and said wider European availability should follow in a few weeks. The Decoder · Engadget · 9to5Mac

• OpenAI said internal models solved three open questions in math, computer science and physics, with outside experts checking the still-unverified proofs. The Decoder

• OpenAI capped paid ChatGPT accounts per device and payment card, blocking new subscriptions or device access when limits are reached; logging out does not automatically free a spot. The Decoder

• OpenAI reorganized its Model Spec around five core behavioral values, including faithfulness to human intent and correctability, with guidance to infer intent from conversation history and disclose assumptions. OpenAI News · The Decoder

• OpenAI's Aura Safety Check can block websites from ChatGPT browsing, according to a report. Search Engine Journal

Developer tools

• Python SDK releases 3.40.0, 3.41.0 and 3.42.0 added cache breakpoint TTL options, Realtime references and reconnection improvements, agent-session controls, image-edit markup overlays and external key management. openai-python releases · openai-python releases · openai-python releases

• Android Studio's Bring Your Own Agent feature added support for more than 40 coding agents, including Codex and Claude Agent. The Decoder

• Codex 0.166.0 added transcript navigation, search, configurable shortcuts and GPT-6.1 Terra entries in bundled, Vertex AI and Amazon Bedrock model catalogs. Codex releases

• Rosalind-6 reportedly costs 1.5 times Astra despite using the same weights. FourWeekMBA

Security and reliability

• Researchers reportedly exploited an OpenAI API flaw to extract hidden GPT-5.6 reasoning traces. CyberSecurityNews

• OpenAI applied a mitigation for degraded GPT-6.1 Sol performance affecting Batch, Chat Completions and Responses and continued monitoring recovery. OpenAI status

• OpenAI mitigated elevated errors affecting most hosted Code Reviews and Security Reviews on private GitHub repositories and continued monitoring recovery. OpenAI status

• OpenAI investigated elevated search errors affecting Dots, ChatGPT Work and some ChatGPT thinking experiences. OpenAI status

• OpenAI mitigated elevated GPT Image 2 API image-edit errors and continued monitoring recovery. OpenAI status

• OpenAI marked elevated errors on POST /ads/* endpoints as resolved, with all affected services fully recovered. OpenAI status

Enterprise deployments

• Fanatics used OpenAI for Shop the Look across NFL, NBA and MLB stores, with testing across roughly 30,000 MLB products showing 10% higher engagement. OpenAI News · Benzinga

• OpenAI and ServiceNow expanded their enterprise workflow partnership. TradingView

• OpenAI said its models help power Retell's 100 million monthly calls, with GPT-6 Astra topping Retell's voice-agent benchmark and Codex helping it reach a $1.6 million annual revenue run rate per employee. OpenAI News

• Legora described using OpenAI models for contract analysis, due diligence and agent-driven legal workflows reviewed by humans. OpenAI News · OpenAI

• EXL became an OpenAI Select Partner to expand enterprise AI deployments. Yahoo Finance · citybiz

Compute supply

• OpenAI reportedly plans to buy Google's TPU chips. BetaNews

• Sam Altman visited Samsung for an AI seminar and meetings reportedly concerning memory supply for future AI data centers. Tom's Hardware · TechPowerUp

What to watch

• Watch for the separate Daybreak Blue training opt-out before the planned November 29 start of session-data training.

• Watch for the December merger of Chat and Work and the introduction of shared ChatGPT–Codex token quotas.

• Watch for wider European Dot availability in a few weeks, as indicated at the Berlin DevDay Exchange.

• Watch for outside experts' verdicts on the three claimed research solutions; timing has not been announced.