ChatGPT news & updates
One ongoing page for ChatGPT, OpenAI models and related developer tools. Verified developments are added to the timeline below, with the sources and context needed to follow what changes.
2 substantive entries added. Coverage: 17 October 2026 at 12:08 BST–18 October 2026 at 12:09 BST. All historical entries and existing visuals remain on this page. Next scheduled scan: 19 October 2026 at 12:00 BST · Europe/London.
The ongoing timeline
Newest update firstHistory stays here
Replit retains GPT and Claude choices in Agent
Replit prefers OpenAI, its CEO clarified. Core and Pro users can select GPT or Claude models in Power or Max mode, according to Replit’s documentation. Free mode is automatic; manual choices vary by rollout, organisation settings and authorisation.
Why it matters: Default placement and model availability are different; check your account’s selector for access.
CEO clarification dated 17 October 2026. This concerns an existing integration, not a new underlying model release.
GPT-6 Astra joins Perplexity’s documented Agent API models
Perplexity has added openai/gpt-6-astra to its documented Agent API choices. Its model reference lists flex and priority service tiers; this is third-party API availability, not a new Astra release or a change to ChatGPT plans.
Why it matters: Developers using Perplexity can evaluate Astra through that provider’s API and billing terms.
Perplexity’s official RSS dates the notice to 17 October 2026 at 12:16 BST. The first callable moment is not separately specified.
OpenAI opens prompt-designed GPT Live voices to all API customers
OpenAI’s 16 October API release note makes prompt-to-voice generally available to API customers: a written description can create an adult synthetic voice without supplying a reference recording. These prompt-generated voices work with GPT Live, not the separate Text-to-Speech, Realtime or Chat Completions APIs; recording-based custom voices remain a different, eligibility-controlled workflow. Creating the voice is free, while preview and subsequent speech usage follow normal GPT Live billing. The same release adds nine built-in voices across German, French, Hindi, Japanese and Korean.
Why it matters: Developers gain another way to design spoken interfaces, but the supported API and the distinction from voice cloning affect integration and consent requirements.
Official release dated 16 October 2026; no precise publication hour is supplied. This is API availability, not a claim that every ChatGPT account received a new voice picker.
| Route | Scope and requirement |
|---|---|
| Written description | API customers; generated voice is GPT Live-only. |
| Reference recording | Separate eligible-customer workflow with consent and sample recordings. |
| Cost boundary | Prompt creation is free; speech use and previews are billed normally. |
Simons launches joint AI-mathematics support
Simons announced $15 million in combined funding and resources with OpenAI and Anthropic for mathematics. The foundation will administer awards for workshops, evaluation of AI-assisted results and shared standards; this is not proof certification.
Why it matters: Community-administered support can strengthen independent evaluation of AI-assisted mathematics.
Announced 15 October 2026 at 15:00 BST; the total combines all three partners.
Cloudflare and OpenAI publish early results from their search-discovery pilot
Cloudflare and OpenAI say their July-started research pilot helped OpenAI discover and index hundreds of millions of previously unknown webpages using signals from a participating cohort of roughly 1% of Cloudflare’s network. The co-authored update describes search and discovery metadata, not sharing page contents for model training. A signal does not grant access permission, and the reported indexing results do not independently establish better answer accuracy or increased publisher revenue.
Why it matters: Finding useful pages is distinct from training on them, and site access choices remain relevant even when discovery signals improve.
Results published 15 October 2026 at 14:00 BST; the pilot began in July. These are results reported by the participating companies, not an independent performance audit.
OpenAI documents native Decisions integration for GPT-Live
OpenAI’s 15 October API notes add a native connection between GPT-Live and the Decisions API. Live can delegate directly for typed decisions, while applications can also request decisions from incoming or outgoing transcript fragments instead of waiting for a complete conversational turn. The application still executes actions and reports their actual results; the integration is not evidence that a spoken commitment completed a task.
Why it matters: Developers gain another way to connect streaming voice to bounded application decisions, while permissions, action execution and verification remain application responsibilities.
Release note dated 15 October 2026; no publication time is supplied. Newly confirmed in this daily scan. The Decisions API’s original beta release on 6 October is a separate event.
| Component | Documented role |
|---|---|
| GPT-Live or application trigger | Select when a decision is requested during the voice session |
| Decisions API | Return a typed answer to a bounded question |
| Application | Execute permitted actions, verify the actual state and report the result |
Eight states announce separate OpenAI cases over alleged ChatGPT safety failures
Eight US states announced individual state-court lawsuits against OpenAI over claims that ChatGPT’s design and safety marketing breached consumer-protection laws. Iowa’s published petition seeks penalties, reimbursement and restrictions on some data practices and advanced-model testing, while Arizona’s attorney general confirms a filing in Maricopa County Superior Court. These are allegations and requested remedies, not judicial findings or restrictions already imposed. Reuters reported that OpenAI did not immediately respond to a request for comment.
Why it matters: The filings expand state scrutiny of chatbot design and marketing, without establishing liability or the outcome of any case.
Petition and announcements dated 14 October 2026; Reuters’ report was updated that evening. This entry confirms the legal actions, not the allegations.
GPT-5.5 reaches its documented Codex cutoff, while API-key access stays outside the change
OpenAI’s current release record makes GPT-5.5’s retirement from Codex effective at 20:00 BST on 14 October for users signed in with ChatGPT. The notice explicitly excludes the OpenAI API and Codex sessions authenticated with an API key. It recommends GPT-6.1 Sol on eligible paid plans and GPT-6 Luna for Free or Go desktop users when those replacements are available to their account and client. Saved defaults, custom agents and scheduled workflows still naming gpt-5.5 need an available replacement; this documents the policy, not a test of every client’s rollout.
Why it matters: Product model catalogues and API availability can differ, so a signed-in Codex cutoff should not be read as an API shutdown.
Effective event: 14 October 2026 at 20:00 BST (19:00 UTC). The current changelog entry is dated 14 October; the earlier September notice was prospective.
| Access route | Documented effect |
|---|---|
| Codex with ChatGPT sign-in | Retirement effective 14 October 2026 at 20:00 BST |
| OpenAI API | Not affected by this notice |
| Codex with an API key | Not affected by this notice |
OpenAI Foundation and Rockefeller launch a $50 million city-services grant call
The OpenAI Foundation and Rockefeller Foundation launched Cities Deliver, a $50 million initiative for eligible organisations working with city governments on AI-supported public services. Planned grants of $500,000–$2 million cover public health, economic mobility, environmental resilience and civic engagement, with initial applications closing on 29 November 2026. The call requires a city-government partner and a meaningful frontier-AI component but does not require or favour OpenAI technology. This is a funding call, not $50 million already paid out or proof that the proposed services work.
Why it matters: The programme creates a concrete funding route for public-service AI while leaving model choice open and future impact to be demonstrated.
Call opened 14 October 2026; Rockefeller’s release was published at 14:38 BST. Initial submissions close 29 November at 23:59 Eastern Time.
OpenAI schedules selected GPT-4 API models for January retirement
OpenAI’s 14 October notice schedules selected GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini API snapshots to shut down on 14 January 2027. Fine-tuned versions follow the same deadline, and the documented migration target is GPT-6 Luna. This is advance notice of an API shutdown, not confirmation that the affected models have already disappeared or a new ChatGPT model launch.
Why it matters: Developers have a concrete migration deadline; they should test replacements against their own workflows rather than assume identical behaviour.
Notice dated 14 October 2026; no publication time is supplied. Scheduled API shutdown: 14 January 2027.
| Model | Snapshot dates |
|---|---|
| GPT-4o | 2024-08-06; 2024-11-20 |
| GPT-4o mini | 2024-07-18 |
| GPT-4.1 | 2025-04-14 |
| GPT-4.1 mini | 2025-04-14 |
All rows: shutdown scheduled for 14 January 2027; recommended replacement gpt-6-luna. Fine-tunes inherit the underlying model’s deadline. Source: OpenAI’s linked deprecation record.
OpenAI plans a Dublin Applied AI Lab and 200 additional Irish roles
OpenAI announced 200 additional planned jobs in Ireland on 12 October, including engineers for a new Applied AI Lab at its EU headquarters. IDA Ireland says the roles add to the 250 announced in July, bringing the combined commitment to at least 450 new roles rather than 450 people already employed. The lab is intended to help customers build and deploy AI applications, and the new Dublin headquarters is due to be occupied from December 2026.
Why it matters: The commitment would add local deployment and privacy-engineering support for European customers; it is a hiring and lab plan, not a new ChatGPT model or a completed expansion.
Announcement: 12 October 2026; Irish Times report: 12 October at 16:35 BST. The primary announcement does not give an event time.
GPT-6.1 Sol gains an Ultrafast option in Work, Codex and the API
OpenAI added Ultrafast processing for GPT-6.1 Sol on 8 October, without introducing a separate model. In Codex and ChatGPT Work, access is restricted to Pro $500 and eligible Enterprise and Edu plans; Enterprise owners must enable it. API developers can request the tier through the Responses API, subject to separate limits, with prices six times the Standard tier and no Free-tier support. The documentation supports US and EU inference residency.
Why it matters: Teams can choose faster token generation, but need to check eligibility, residency and the higher usage cost.
Independent reports put OpenAI’s September revenue run rate near $50 billion
CNBC and Bloomberg reported that OpenAI’s annualized revenue run rate was roughly $50 billion at the end of September, citing an investor update and people familiar with the figures. The reporting distinguishes that measure from an earlier gross figure that included partner revenue. A run rate extrapolates recent revenue; it is not audited annual revenue or profit.
Why it matters: The accounting distinction matters when assessing the commercial scale of the company behind ChatGPT.
USA TODAY brings a new copyright case against OpenAI
USA TODAY and thirteen affiliated companies filed a federal copyright complaint against OpenAI in New York on 8 October. They allege unauthorized use of journalism from nineteen publications for model training and in chatbot responses, and seek damages exceeding $250 million alongside other remedies. These are the plaintiffs’ allegations, not findings by the court.
Why it matters: The filing adds a major publisher’s claims to the continuing litigation over AI training and news content.
GPT-6 and Intelligent UI begin rolling out in ChatGPT
OpenAI began rolling out GPT-6 in ChatGPT with answers that combine text, visuals and interactive tools, and can begin appearing while reasoning or tool use continues. The global rollout started on 7 October for Plus, Pro, Business and Enterprise, with expansion to Free and Go announced for 8 October; Enterprise access depends on administrator settings. Intelligent UI works in Chat from Instant through Extra High reasoning, but is unavailable in the Pro reasoning option, Voice and Work.
Why it matters: ChatGPT can present everyday explanations and planning tasks as interactive experiences that people can use inside the conversation.
Audio uploads arrive in ChatGPT
Paid ChatGPT users can attach supported audio recordings and request transcripts, summaries or answers about their contents. Audio uploads are available with paid subscriptions and workspaces, including Enterprise, but are unavailable on Free; access can vary by region, client, model and workspace settings. OpenAI cautions that transcripts can contain errors and speaker identification may be unreliable.
Why it matters: Meetings, interviews and lectures can become source material for a normal ChatGPT conversation without a separate transcription step.