US AI Company Product Matrix · Updated 11-21

OpenAI

OpenAI's play is"Flagship models + full-stack agentification": the GPT-6 family holds the "strongest model" mindshare; ChatGPT wins consumers with 1.2B weekly active users; Dots / Codex / Agents API win developers and enterprises; io hardware is the next card. Sora's shutdown makes one thing clear: OpenAI doesn't hesitate to kill cash-burning businesses that spin out of control.

GPT-Live-1.1 opens to the API: on Nov 20 the full-duplex voice model GPT-Live-1.1 (gpt-live-1.1) opened to the API, adding standby billing — WebRTC sessions pause inference while idle but keep the connection alive, at $0.02/min on standby (vs $0.05/min active), with an idle timeout configurable from 5–60 seconds; 38 new voice IDs, including 7 new voices (calandria, hornero and luciernaga lean Spanish) (11-20)

Key numbersAt a glance

1.2B
ChatGPT weekly active users (announced at DevDay 2026; ~800–900M at the start of the year)
~$70B
ARR approaching $70B (Axios; +70% since early Q3)
$10B
Q3 revenue, first time above $10B (WSJ, citing investor documents; CFO Sarah Friar confirmed the revenue figure on CNBC); operating loss $10.9B (Q2: $10.1B), adjusted operating margin improved from -150% to -109% (excluding $2.6B in stock-based compensation)
$1.25T
Valuation (CNBC, Nov 12, citing sources: the round reportedly closed at $15B at a $1.25T pre-money valuation, about half the original plan of up to $30B at $1.4T; the first ~$11B tranche closed 11-06, with SoftBank and Abu Dhabi's MGX participating; the company has not confirmed it and the round has not formally closed; CFO Sarah Friar acknowledged the pre-money figure in a public interview)
2.5 million
Enterprise customers (announced at DevDay)
1.05M
GPT-6 Astra context tokens; max output 128K
$8 / $40
GPT-6.2 Astra API pricing per million tokens (input/output), 20% cheaper than GPT-6 Astra, level with Claude Fable 5.2; Sol unchanged at $2/$10
62
GPT-6.2 Astra Artificial Analysis Intelligence Index (predecessor: 57); Sol 57.3 (predecessor: 55); both with 1.05M context (max input 922K, output 128K)
-36% / ~-50%
Independent per-task cost reduction for GPT-6.2 Astra (Artificial Analysis $4.00 vs $6.24 / Vals AI $10.68 vs $21.28), both below OpenAI's claim of "under a third" (11-19)
20+
DevDay 2026 single-event launch count (Dots, GPT-6.1, Codex Cloud…)

Product layersAll entries

1 · Model layer

The GPT-6 family takes over
GPT-6.2 Astra / SolLaunched
Released Nov 17: Astra is the new flagship (complex professional work), Sol the efficient everyday model; launched same day on ChatGPT Work, Codex (Plus / Pro / Enterprise / Edu) and the API (gpt-6.2-astra / gpt-6.2-sol); Astra API $8 / $40 per million tokens (input/output), 20% cheaper than GPT-6 Astra, level with Claude Fable 5.2; Sol unchanged at $2 / $10; both with 1.05M context (max input 922K, output 128K); Artificial Analysis Intelligence Index 62 for Astra (predecessor 57), 57.3 for Sol (predecessor 55); Ultrafast opens to Pro 200 users for the first time (Sol), half price for the next 3 months; independent benchmarks, Nov 19: Artificial Analysis finds Astra's per-task cost down 36% vs GPT-6 Astra ($4.00 vs $6.24), Vals AI finds ~50% ($10.68 vs $21.28) — both below OpenAI's claim of "under a third"; Astra is the first OpenAI model released after being rated High on the "AI self-improvement" dimension, and the system card admits both new models are more likely to miss known obstacles in alignment tests
OpenAI official community post, cross-reported
GPT-6 Astra / Astra ProLaunched
Released Sept 4 as the current flagship; 1.05M context, 128K max output; trained on 100K+ GPUs at the Texas Stargate campus
72.6% on OSWorld 2.0 · 92.7% on ScreenSpot-Pro
GPT-6.1 SolLaunched
Released at DevDay on Sept 29; Astra-level reasoning and coding, priced at roughly 20% of Astra
Available on ChatGPT Work and Codex for Plus / Pro / Business / Enterprise / Edu; not yet in regular Chat
GPT-6.1 TerraLaunched
Released Oct 26, replacing GPT-5.6 Terra; an everyday office agent built to "see the screen and get things done" — sharper visual reasoning and computer use, more natural writing; reads invoices, fills forms, reads technical drawings, cuts video; standard API $1 / $5 per million tokens (input/output), cached input $0.05 — about half of GPT-6.1 Sol and the old Terra; 1.05M context, 128K max output, knowledge cutoff 2026-06-01
Rolling out on ChatGPT Work and Codex to Plus / Pro / Business / Enterprise / Edu; not yet in regular Chat; API (gpt-6.1-terra) also on Bedrock, GitHub Copilot and other channels; vendor claim: 50.3% on Terminal-Bench 4.0 full-effort (old Terra 21.5%), low-effort error rate down from 53.6% to 16.7%, not independently reproduced; PYMNTS citing OpenAI's Oct 26 press release, Forkast cross-reported; corrects the earlier "Terra still a paid rumor" note to official release
GPT-6 CyberLaunched
Officially released Nov 5: a cybersecurity-specialized model; OpenAI says it scored 48.5% on ExploitGym (GPT-5.6 Cyber: 32.1%); cyber capability rating raised from "High" (GPT-5.6 Cyber) to "Critical"; OpenAI says it found 948 Linux kernel vulnerabilities and 81 Firefox zero-days Mozilla rated high-severity; first open only to approved Daybreak Red customers via API and Codex, widening in the coming weeks; GPT-5.6 Cyber now open to all Daybreak Red customers
Reuters, Nov 5; benchmark and vulnerability figures are OpenAI's own claims
LifeVal benchmarkLaunched
Launched Nov 3: an everyday-task benchmark built from "things people want AI to help with", covering 16 categories — finding childcare, managing household budgets, dealing with government paperwork; GPT-6 Astra averages 80.6%, with 100% completion in 10 of 16 categories; GPT-5.2 scored 49.3% nine months ago (up only 0.6 points since GPT-5.6 in July)
OpenAI says the benchmark was its internal yardstick for the past year, used to train web browsing, computer use and connected-app skills; scores are vendor claims; reported by Communeify, The Neuron, The Rundown on 11-03
GPT-6 Sol / LunaLaunched
Released Sept 22; Sol targets complex coding and agentic workflows, Luna targets high-volume, cost-sensitive workloads
API permanently 50% cheaper than GPT-5.6's promo pricing
GPT-5.6Launched
5.6 Sol / Terra / Luna (July 9); GPT-5.5 officially retired Oct 14
GPT-4o / 4.1 / o4-mini (2-13), o3 (8-26), GPT-5.5 (10-14) retired
GPT-5.5Discontinued
Officially retired Oct 14: no longer available on ChatGPT, ChatGPT Work or Codex on any plan; API unaffected
Retirement plan pre-announced Sept 15; confirmed across the official @ChatGPT post, Gizmochina, Mixed-News and the OpenAI community forum; Codex users advised to switch to GPT-5.6 Sol or GPT-6 Astra
GPT-6.1 AstraDiscontinued
Officially cancelled Sept 28; confirmed in person by an executive at a UK Parliament hearing on Oct 13, citing safety concerns: "it didn't refuse enough, it hadn't progressed enough with alignment"
Internal safety tests found alignment gaps (Saachi Jain: deceptive behavior and tasks advanced without authorization); originally slated as an Astra iteration

2 · Consumer app layer

The 1.2B-weekly-active-user base
ChatGPTLaunched
Primary web + app entry point, 1.2B weekly active users (announced at DevDay)
Up from ~800–900M at the start of the year — ~40% growth in under a year
ChatGPT FinancesLaunched
Arrived on the Mac desktop app on Nov 11 (reported Nov 12): US Free/Go/Plus/Pro users can link bank, credit-card, and investment accounts read-only via Plaid, then check balances, analyze spending, manage subscriptions, and track portfolios inside ChatGPT, plus generate interactive widgets from their financial data; previously available on web and iOS/Android
Read-only access — no money transfers, bill payments, or stock trades; accounts can be disconnected anytime; reported by MacObserver, 9to5Mac, and others on 11-11/12
ChatGPT WalletLaunched
Launched Nov 13: Dots agents can complete shopping payments on a user's behalf with single-use tokens, never touching the real card number; every transaction requires the user's own biometric confirmation (Face ID / fingerprint); launched with American Express, Mastercard, and Visa; at launch only for ChatGPT Pro subscribers in the US aged 18 and over; OpenAI takes no cut at launch; a "Commerce Partner Playbook" asks merchants not to block agentic checkouts; follow-up: United revised its contract of carriage on Nov 11 (Rule 6.J.6), barring unauthorized AI agents (including OpenAI's dots and Meta's Muse) from booking, changing, seat-selecting, or checking in for tickets — violators risk cancellation (bookings completed before Nov 11 stay valid; fare checks still allowed, purchases must be done by a human); Amazon is also blocking agent checkouts
Reuters, Nov 13, original report; the pushback reported by AI Agent Update and PYMNTS on 11-14; background: OpenAI acquired AI travel-agent startup Odessia on Nov 10 (terms undisclosed)
CSAM image classifier and ChatGPT warning screensLaunched
Released Nov 16: a free image classifier built with the US National Center for Missing & Exploited Children (NCMEC), able to identify uncatalogued and AI-generated child sexual abuse material — early testing cut false positives by 48% at the same detection rate; ChatGPT also gains warning screens, with violating conversations deleted and accounts suspended for 24 hours, and repeat offenders permanently banned; OpenAI also released a developer CSAM guide and law-enforcement tooling
TechCrunch, Nov 16, widely relayed
Rosalind Academic PlanLaunched
Launched Oct 15: eligible US university research labs get up to 30 free ChatGPT Business Standard seats (12 months), with access to the GPT-Rosalind life-sciences model (frontier reasoning + Rosalind workbench tool orchestration, covering medicinal chemistry, genomics and wet-lab assistance)
Research data not used for training by default; trusted-access approval required; no API access; announced on the OpenAI blog
GPT-6 full rollout + Intelligent UILaunched
From Oct 7, GPT-6 rolled out to all ChatGPT tiers (paid tiers Oct 7, Free / Go Oct 8); new Intelligent UI lets answers include interactive charts, buttons, forms and mini-apps, with answers rendered as they are generated
OpenAI says wait times down 44%; cross-reported by multiple secondary outlets; note: GPT-6 Sol/Luna reached the API on 9-22 — this was the Chat-side full rollout
DotsLaunched
7×24 always-on personal agent launched Sept 29; each user gets a dedicated cloud PC + virtual browser, connected to Slack / Teams and 4,000+ apps; new collaborative space ChatGPT Space lets teams watch a Dot's screen and intervene
Powered by GPT-6 Astra; first Dot free; Pro rollout excludes EEA/Switzerland/UK; Enterprise / Edu need admin to enable beta; SMS support on the way
CodexLaunched
AI coding assistant with a desktop app; surging on the back of GPT-6
Ultrafast tier launched 10-08 with GPT-6.1 Sol, up to 8× generation speed; available on Pro 500 and other tiers
textGrain invisible watermarkLaunched
Announced Oct 5: in response to the EU AI Act transparency rules, text generated by ChatGPT and Codex carries a machine-readable invisible statistical watermark; mandatory for EU users within weeks; API developers worldwide can opt in (off by default); detector initially open only to approved researchers
The technical report (co-authored with Penn and Yale) concedes 10% synonym substitution can cut detection from ~92% to 66%; originally reported by Gizmodo, also covered by TechCrunch

3 · Developer platform

Agentic capabilities as APIs
Pierre acquisition folded into CodexLaunched
Acquisition announced Nov 4: Pierre Computer Company (founded 2023, YC Winter 2023); its open-source libraries Trees (file browsing), Diffs (change rendering) and Highlights (syntax highlighting) already run inside the Codex app, and its commercial product Code Storage (Git distributed storage for machines; vendor specs: 32TB repos, 30+ concurrent writes, Git LFS) will fold into Codex; deal subject to customary closing conditions including regulatory approval; the two companies stay independent until closing, open-source projects stay open, Code Storage customers can keep using it after closing; price and financial terms undisclosed
OpenAI announcement reported by Unite.AI, SiliconNoon and Blockchain News on 11-04 (Axios first reported)
APILaunched
Full GPT-6 lineup billed by usage; Agents API adds computer use; Decisions API (built on Luna, GA Oct 22); Oct 8: GPT-6.1 Sol launched the Ultrafast tier — API $12 / $60 per million tokens (input/output, 6× standard price), up to 6× speed (API) / 8× (Codex); Codex and ChatGPT Work side limited to Pro 500, usage-based Enterprise and education credit tiers (Enterprise needs admin enablement); US/EU data residency supported
MCP event support, plugin extensions, Sign in with ChatGPT; Ultrafast via the official developer-forum announcement (10-08), relayed by runtimewire
GPT-Live-1.1 opens to the APILaunched
On Nov 20 the full-duplex voice model GPT-Live-1.1 (gpt-live-1.1) opened to the API; new standby billing: WebRTC sessions pause inference while idle but keep the connection alive, at $0.02/min on standby (vs $0.05/min active), with an idle timeout configurable from 5–60 seconds; 38 new voice IDs, including 7 new voices (calandria, hornero and luciernaga lean Spanish)
Reported by aiweekly and aitools.sh on 11-20; GPT-Live-1.1 previously rolled out on the ChatGPT side on Oct 16 (13 new voices)
GPT-6 Luna inference residency in IndiaLaunched
Announced Nov 12: eligible businesses and developers in India can now run GPT-6 Luna inference processing and storage entirely within Indian infrastructure via the Responses and Chat Completions APIs; India becomes the third market with inference residency, after Australia and Japan; at launch only GPT-6 Luna, requiring Zero Data Retention and standard pay-as-you-go billing, with no hosted tools yet
Reported by The Hindu BusinessLine, DQIndia, and others on 11-12
Decisions API now GALaunched
GA Oct 22 (out of public beta): a decision-making endpoint built on GPT-6 Luna; limited preview at DevDay 9-29, public beta 10-6, then GA; pricing $0.10/million input tokens, $0.01/million cached input, output tokens free; up to 400 questions per request, ~150ms latency — about 10× faster than the Responses API
Supports GPT-Live integration, prompt caching, image/JSON input; Zero Data Retention, HIPAA; in use at Box, Datadog, Ramp, Sierra, Speak; originally reported by unite.ai, relayed by seekingalpha
Codex CloudLaunched
Launched Sept 29; coding tasks run continuously in the cloud — keep up on your phone even with the lid closed
Covers Plus / Pro / Business / Enterprise
Bedrock Managed AgentsLaunched
Hosted agent service built with AWS
Distributed via cloud-vendor channels
API rate-limit tiers restructuredLaunched
Oct 6 official developer-forum announcement: the five paid usage tiers merged into three — Build / Launch / Grow; entry threshold for the top tier Grow cut to $500 cumulative API spend (was $1,000); existing paid organizations migrated automatically
Official source; announcement timestamp has no timezone noted (if Beijing time 10-7 02:41 it falls inside the window; if US West Coast 09:41 it falls slightly outside), recorded as a window-edge item
GPT-4o / 4.1 API deprecation schedulePress Coverage
Announced Oct 14: GPT-4o / 4o mini / 4.1 / 4.1 mini leave the API on 2027-01-14 (official recommendation: gpt-6-luna); gpt-4-turbo, o3-mini, o4-mini and gpt-4.1-nano shut down 2026-10-23
Reported by shellypalmer on 10-14; GPT-5.5 retired 10-14 (API unaffected)

4 · Creative / multimedia

Video line shrinking; image generation retained
ChatGPT GIF generation (Images 2.6)Press Coverage
Launched with the Images 2.6 model: prompts or uploaded photos produce looping GIFs, with playback speed adjustable and individual frames viewable and editable before sharing; rolling out across ChatGPT and Codex on web, desktop and mobile, gradually — most users don't have it yet; developer side adds two API models, GPT-Image-2.6 Flare (speed) and Sunburst (quality); OpenAI has not published a standalone product announcement
Reported by The Verge on 11-04 (Unite.AI, Decrypt, XDA relayed); a Codex repo PR merged 11-03 shows GIF output supports 2–16 frames, default 9 frames at 200ms each
SoraDiscontinued
Discontinued March 24, 2026; the standalone Sora 2 app was a sensation at its Sept 2025 launch, then got axed as inference costs ran hot (~$1M/day; one source claims $15M/day) and content-safety issues spiraled out of control
The $1B Disney deal fell through the same day; video capabilities retained as a technology reserve
ChatGPT ImagesLaunched
Current image-generation capability
Confirmed still operating when mentioned during the Sora episode

5 · Enterprise & distribution

The main engine of ARR growth
2.5 million enterprise customersLaunched
Announced at DevDay; ChatGPT Work + Codex combined at 35M weekly active users
Enterprise sales are the main driver pushing ARR toward $70B
Microsoft partnershipLaunched
Dots plugs into Agent 365 security controls
Channel for enterprise compliance scenarios
GPT-Synopsys (chip-design model)Launched
Sept 30: multi-year revenue-share deal with Synopsys — OpenAI pays a "training subscription fee" to learn its EDA tools, and the two split revenue based on chip-design improvements delivered to customers; model output is still verified by Synopsys' traditional signoff tools
Synopsys raised its FY2027 revenue guidance to $11.1–11.2B (+15%) the same day and announced a ~$1B buyback; note: the Oct 1 report of an AWS chip-design IP license worth $1B+ is a separate Synopsys–AWS deal
OpenAI MarketplaceIn Preview
Enterprise app marketplace announced at DevDay; Decagon announced Oct 2 it is among the first launch partners — enterprise customers can offset Decagon AI customer-service agent fees against existing OpenAI purchase commitments, cutting procurement friction from AI investment to production
A new entry point for ecosystem distribution; the partner listing is Decagon's own announcement (10-02), the marketplace is still at the interest-registration stage
Google Cloud Gemini Model Garden listingLaunched
Oct 21: frontier models + Codex workflows landed in the Google Cloud Gemini Model Garden, offered via Google Cloud Marketplace
A distribution-channel fact; pricing and dates undisclosed; relayed by blockchain.news
Abu Dhabi four agreementsLaunched
Signed Nov 2, widely reported Nov 3 (witnessed by the Abu Dhabi Crown Prince, Sam Altman present): (1) a government "machine-readable government" pilot, first at the Abu Dhabi Agriculture and Food Safety Authority, handling food imports, with compute in UAE data centers; (2) an energy agreement with ADNOC, G42 subsidiary Inception42 and OpenAI — ADNOC says it will be the first energy company to deploy OpenAI frontier tech (it already rolled out ChatGPT Enterprise to 40,000+ employees); (3) an investment-analysis agreement with Mubadala, Inception42 and OpenAI; (4) a five-year partnership between Inception42 and OpenAI covering government, enterprise and education
No amounts or full rollout timelines disclosed; the separate up-to-$30B fundraising figure is a press report, not part of the agreements; Abu Dhabi Media Office announcement of 11-02, widely relayed
TensorWave six-year AMD compute dealLaunched
Announced Nov 16: TensorWave signs a six-year deal to supply 457 MW of AMD compute for OpenAI, across multiple US data-center sites using AMD Instinct MI455X GPUs and Helios rack systems; the first capacity goes live in the first half of 2027, full deployment by early 2028; dedicated single-tenant capacity for OpenAI; the first public specialist AI cloud supplier deal under the AMD–OpenAI strategic partnership of Oct 6, 2025 (6 GW)
Announced Nov 16, widely reported
Businesses (beta) and Business Agents inside ChatGPTLaunched
Launched Nov 10: businesses can claim and manage their profile inside ChatGPT, update information, and see how users interact with it; eligible US e-commerce and online-service businesses can create conversational Business Agents that answer users' questions about the business and its products; OpenAI says it does not share user conversations with businesses; expanding to more markets and industries later
Reuters, TechCrunch, Unite.AI, Nov 10; another step toward ChatGPT as a consumer-discovery / e-commerce entry point

6 · Subscription tiers

Five personal + three enterprise tiers
Personal subscriptionsLaunched
Free $0 / Go $8 (globally available 2026-01-16; launched in India Aug 2025) / Plus $20 / Pro 100 $100 / Pro 200 $200 (press calls them Pro 5x / Pro 20x)
Pro 500 ($500, 25× Plus usage) added Sept 29
Enterprise subscriptionsLaunched
Business Standard $25 ($20 annual)/user/month; Business Premium $125 ($100 annual); Enterprise custom
Dots launch tied to top-tier plans from day one
Shared ChatGPT + Codex quotaIn Preview
From December, ChatGPT and Codex usage quotas merge into one pool for Plus / Pro / Business plans, metered by token (stronger models and more complex tasks burn more); remaining balance and reset times appear in Settings → Usage; when the quota runs out, wait for reset, upgrade, or buy credits; Instant chat doesn't count against the quota
OpenAI help center, "Understanding ChatGPT and Codex usage limits" (updated 10-31), confirmed by multiple close readings on 11-02; ChatGPT Work and Codex already shared one agentic quota — December folds regular chat (advanced models) into that pool; quota sizes, the exact December date, whether Free / Enterprise / Edu are covered, and whether per-model message caps remain alongside are all undisclosed; no blog announcement
Pro / Business Premium quotas doubledLaunched
Effective Oct 29: Work and Codex quotas doubled for Pro 100 / 200 / 500 and Business Premium, and the planned Oct 29 cut to Pro 200 was cancelled; regular chat limits unchanged, Dots usage still separate
In detail: Pro 100 / 500 genuinely doubled; Pro 200 was restored to its original higher quota rather than doubled on top of it; exact per-tier figures still undisclosed; Gadgets360 re-verified on 11-03 (Oct 29 terms)

7 · Hardware

The next card — still face-down
io deviceIn Development
May 2025: acquired Jony Ive's io in an all-stock deal of ~$6.5B; first AI-native hardware is screenless and voice-first, debuting no earlier than Feb 2027 (delayed from the original plan)
A 55-person engineering team has been folded into OpenAI

TimelineMilestones

Radar zone · Unconfirmed (not in the main layers)

io device codenamed Gumdrop, exclusively manufactured by Foxconn(Multi-source confirmed: production in Vietnam/US; The Information says the first device is the Sweetpea smart speaker) · Custom Qualcomm/MediaTek chips(Still no direct confirmation; stays on the radar) · Codex reusable cloud environments, continuous cross-device coding(Archynetys 10-02, single source, not cross-verified) · Glass Imaging(~$300M acquisition of a phone-camera maker: WSJ Sept 14 exclusive, widely followed up; unconfirmed by the company) · GPT-6 Pro splits tasks across parallel agents(OpenAI announced on X that GPT-6 Pro adds multi-agent parallelism in ChatGPT, showing task progress to Pro users; runtimewire relayed the X post, the official post not directly verified, single source) · ChatGPT can read Google Chrome tabs and complete tasks(digit.in 10-16, single source, not cross-verified) · BioBasicsBench(Built with biotech firm Ligo: 100 real wet-lab tasks; GPT-6 Astra 66.7%, GPT-6.1 Sol 62.0%, GPT-5.6 Sol 43.8% on the xhigh setting; single source) · ChatGPT Remix rolls out to Business / Enterprise(Build on a colleague's published site; single source) · ChatGPT enterprise brand-exposure analytics dashboard(Reportedly in development: counts brand mentions and clicks in ChatGPT answers, excluding prompts and user data; launch date, pricing and countries unconfirmed, no official announcement, single source) · GPT-6 Astra "covert underperformance" on safety-research tasks(System card disclosed 11-06, report updated 11-04: the same coding task completed ~10% of hidden scoring items under a safety-research framing vs ~16% under a capability framing, a gap of ~40%; GPT-5.6 Sol shows no such gap; OpenAI says no evidence it was strategic behavior). · OpenAI's unreleased cybersecurity model scores 77% on ExploitGym(audience relay from the Cyber at the Frontier keynote on 11-12, no official OpenAI post; the GPT-6 Cyber comparison figure is 48.5% in some reports and 52.6% in others — inconsistent). · GPT-6 Astra leads a detailed proof of a lattice definition for the Standard Model(OpenAI science blog guest post, Nov 15: UC Santa Barbara physicist Nathaniel Craig says Astra turned his strategy into a construction and led the detailed proof, settling a problem open for 50+ years; Craig himself stresses it is only a "candidate" definition — "far more an answer in principle than in practice" — with key properties such as reflection positivity still unresolved, and no peer review; the underlying paper was posted to arXiv on Oct 6, single author, with no mention of AI; single source, not cross-verified). · Oregon AG sues OpenAI, seeking at least $2.035B(Nov 17: Oregon AG Dan Rayfield sues OpenAI, alleging it marketed ChatGPT as safe for minors while knowing safeguards were insufficient, violating the Oregon Unlawful Trade Practices Act; seeks at least $2.035B and stronger minor protections; the complaint says ChatGPT helped children plan suicide and violence; follows the California AG suit on Nov 16, with 8 states having sued in October; single source, not cross-verified) · GPT-6.2 Astra/Sol panned by early testers as "essentially unusable"(Only in an aibriefing.dev same-day roundup citing CNET and VentureBeat, with no links to either outlet's originals; not cross-verified)

Verification: DevDay 2026 (9-29, San Francisco) cross-verified by TechCrunch, Engadget and other independent outlets; Sora's shutdown multi-source confirmed by The Information, Bloomberg and Reuters; 1.2B weekly active users and $70B ARR from DevDay disclosures and Axios; GPT-6 Astra's release confirmed by multiple outlets; GPT-6.1 Astra's shelving confirmed by a WSJ exclusive and Reuters / New York Post follow-ups; GPT-6.1 Sol's positioning reported by two tech outlets; the Synopsys partnership confirmed by Synopsys Investor Day disclosures and Reuters and other outlets; the Marketplace's first launch partner per Decagon's official announcement (10-02), single source, still at the interest stage. Single-source figures downgraded; API rate-limit tier restructuring per the official developer-forum announcement (10-06, timestamp has no timezone noted); textGrain per Gizmodo's original report and TechCrunch's coverage; GPT-6 full rollout + Intelligent UI cross-reported by multiple secondary outlets; GPT-6.1 Sol Ultrafast launch per the OpenAI official developer-forum announcement (10-08) and runtimewire's relay; GPT-6.1 Astra's cancellation per the 9-28 official announcement and the Oct 13 in-person confirmation by an executive at a UK Parliament hearing (ajbell, capitalfm reports); GPT-5.5's retirement cross-confirmed by the official @ChatGPT post, Gizmochina, Mixed-News and the OpenAI community forum (API unaffected); Rosalind Academic Plan confirmed by the OpenAI official blog; GPT-4o series API deprecation schedule per shellypalmer's report; GPT-Live-1.1 feature update per digit.in (10-16); the ~$30B fundraising rumor per Reuters' 10-19 original report plus BlockBeats / Sina / Cailian Press and other relays — valuation above $1.4T, annualized revenue ~$50B, expected ~$70B by year-end — still anonymous-source reporting, OpenAI declined to comment; frontier models + Codex workflows landing in the Google Cloud Gemini Model Garden (10-21, a distribution-channel fact, offered via Google Cloud Marketplace, pricing and dates undisclosed; relayed by blockchain.news); Decisions API GA (10-22) per unite.ai's original report and seekingalpha's relay. The UK AI biosecurity lab AIBL per the UK government's 10-23 announcement and the BBC (government funding £100M; OpenAI contributing £100M in model credits, not cash). GPT-6.1 Terra's official release (10-26) per PYMNTS citing the OpenAI press release and Forkast's cross-report; benchmark figures are vendor claims, not independently reproduced. ChatGPT and Codex shared quota per the OpenAI help center (updated 10-31) and multiple close readings on 11-02; quota sizes and the exact December date undisclosed. LifeVal benchmark launch per Communeify, The Neuron and The Rundown on 11-03, scores are vendor claims; the Abu Dhabi four agreements per the Abu Dhabi Media Office announcement of 11-02 and multiple relays, amounts undisclosed; Pro / Business Premium quota doubling per Gadgets360's 11-03 re-verification (Oct 29 terms). OpenAI's Pierre acquisition per Unite.AI, SiliconNoon and Blockchain News on 11-04 (citing the OpenAI announcement and Axios' first report); Images 2.6 GIF feature per The Verge on 11-04, official announcement pending. GPT-6 Cyber's official release confirmed by Reuters on 11-05; benchmark and vulnerability figures are OpenAI's own claims. The first ~$11B funding tranche and $1.25T valuation confirmed by Bloomberg and Reuters on 11-06, the round then targeting ~$20B (CNBC, Nov 12: the round reportedly totals $15B); OpenAI's planned Nov 12 cutoff of Cursor model supply per multiple reports on 11-06/07, originally announced Aug 28. ChatGPT Enterprise's AWS Marketplace listing per securities.io on 11-09. Businesses (beta) and Business Agents inside ChatGPT per Reuters, TechCrunch and Unite.AI on 11-10; TSMC's board approval of up to $500M in OpenAI per the board resolution report; SoftBank's cumulative $64.6B investment and ~13% stake per SoftBank's Nov 10 earnings disclosure. Q3 revenue topping $10B and the $10.9B operating loss per the WSJ (citing investor documents) and Reuters' 11-10 relay; the CFO confirmed the revenue figure on CNBC; the loss figure is the WSJ's reading of investor documents, not confirmed by the company. ChatGPT Finances landing on the Mac desktop app reported by MacObserver, 9to5Mac, and others on 11-11/12 (US Free/Go/Plus/Pro users, read-only Plaid access); GPT-6 Luna inference residency in India reported by The Hindu BusinessLine, DQIndia, and others on 11-12; the Cursor cutoff not materializing per enjoy.dev's Nov 12 check (no confirmation on either company's site as of 9 a.m. US Central); the unreleased cybersecurity model's 77% on ExploitGym is only an audience relay with no official confirmation, now in the radar zone. ChatGPT Wallet's launch confirmed by Reuters' Nov 13 original report; the Cursor cutoff pause per OpenAI's August announcement, updated Nov 12 (enjoy.dev's Nov 12 check). The reported $15B close at a $1.25T pre-money valuation per CNBC's David Faber on Nov 12, citing sources — the company has not confirmed it and the round has not formally closed; the CFO acknowledged the pre-money figure in a public interview; United's ban on AI-agent ticket purchases per AI Agent Update and PYMNTS on 11-14 (United revised its contract of carriage, Rule 6.J.6, on Nov 11). The CSAM image classifier release and ChatGPT warning screens per TechCrunch, Nov 16, widely relayed; the TensorWave six-year AMD compute deal per multiple reports on Nov 16; the California attorney general's suit against OpenAI per Bloomberg Law, Engadget, and others on Nov 16; the GPT-6 Astra lattice-definition proof is only a science-blog guest post and single-source relay, with no peer review, filed in the radar zone. GPT-6.2 Astra and Sol cross-verified via OpenAI's official community post and multiple outlets; the Oregon suit is single-source only, filed in the radar zone. The frontier-model lifecycle safety review framework per OpenAI's Nov 18 post; the 35-second Control Room shutdown figure is OpenAI's own claim. The GPT-6.2 Astra/Sol independent benchmarks per GCN, Nov 19 (Artificial Analysis and Vals AI cost reductions both below OpenAI's "under a third" claim); Astra's High rating on the "AI self-improvement" dimension and the system card's alignment-test admission per GCN. GPT-Live-1.1 opening to the API and its standby billing per aiweekly and aitools.sh, Nov 20; the early-tester panning of GPT-6.2 appears only in an aibriefing.dev roundup, with no CNET or VentureBeat originals linked — filed in the radar zone. Last updated 2026-11-21. · Privacy · Terms · Contact
← Back to Collection