GPT-5.5 / GPT-5.5 Pro, ChatGPT, Codex, Sora, DALLΒ·E.
GPT-5.5/5.5 Pro (Apr 23) plus specialist frontier models (GPT-Rosalind, GPT-5.5-Cyber validated by UK AISI) and GPT-5-class Realtime voice give the broadest frontier portfolio, with GPT-5.5 Instant cutting hallucinations 52.5% on high-stakes prompts. However, it now trails Claude Fable 5 decisively on agentic coding (58.6% vs 80.3% SWE-Bench Pro).β Claude Fable 5
An internal reasoning model autonomously disproved the 80-year-old ErdΕs unit-distance conjecture with a 125-page proof (May 20) β called 'a milestone in AI mathematics' by Tim Gowers β alongside Rosalind and the Dreaming memory initiative. Counterpoint: open research contributions remain thin relative to Google's open-weights releases like DiffusionGemma.β Claude Fable 5
Relentless surface expansion: Codex at 4M+ weekly developers growing into a general enterprise agent (plugins, Sites, Computer Use on Windows), plus ChatGPT verticals in personal finance (Plaid, 12,000 institutions) and job search, unified under Brockman's super-app org. The ads push risks trust friction, and Canvas removal shows some product churn.β Claude Fable 5
Confidential S-1 filed June 8 at an $852Bβ$1T valuation after a $122B round, $25B+ annualized revenue, the $10B Deployment Company JV, and new monetization (ads, agentic commerce). Counterpoints: Anthropic now leads LLM revenue share (31.4% vs 29% in Q1), the CFO admits a persistent compute-affordability gap, and the Florida lawsuit plus Apple friction add pre-IPO overhang.β Claude Fable 5
Declared $1.4T infrastructure commitments spanning Stargate/Oracle plus multi-cloud delivery via AWS Bedrock GA and Oracle OCI. But OpenAI owns little capacity itself and its own CFO describes a compute crunch where demand exceeds what the company can afford β the binding constraint on scaling.β Claude Fable 5
Concrete shipped safeguards β Trusted Access tiers for cyber models, Trusted Contact for suicide-risk signals, Lockdown Mode against prompt-injection exfiltration, and stated military red lines. Yet Florida's first-of-its-kind consumer-safety lawsuit (June 1) and the simultaneous push into ChatGPT advertising cut against the record.β Claude Fable 5
Deep bench with product consolidated under co-founder Brockman and rising stars like Sottiaux, plus 150 forward-deployed engineers via the Tomoro acquisition. The counterpoint is sustained marquee attrition β most recently co-founder Karpathy joining Anthropic β and repeated reorgs that signal organizational churn.β Claude Fable 5
~900M users, GA on AWS Bedrock and Oracle credits, Dell on-prem, Pentagon IL6/IL7 access, Accenture Federal's 15,000 professionals, and the Malta national rollout form an unmatched multi-channel push. The fraying Apple partnership threatens the single most valuable consumer default surface.β Claude Fable 5
The largest consumer usage feedback loop in AI plus new opt-in proprietary signals (Plaid financial accounts, memory/Dreaming personalization) and growing enterprise switching costs via Codex. Still, it lacks a structural first-party data estate like Google's Search/YouTube or Meta's social graph.β Claude Fable 5
Blistering eight-week cadence: GPT-5.5 family, voice stack, ErdΕs result, AWS/Oracle/Dell channels, IPO filing, and new verticals nearly every week. Headwinds are real but secondary: litigation, Apple friction, and Anthropic flipping the enterprise revenue-share narrative.β Claude Fable 5
Tracked for AI Power Rankings scoring. Covers model releases, benchmarks, pricing, funding, partnerships, infrastructure, and policy changes.
Published: 2026-06-11 | Logged: 2026-06-11T00:00Z | Area: Distribution & Reach, Research & Innovation
OpenAI published a case study on how astrophysicist Chi-kwan Chan uses Codex to build and refine black hole simulations used to study extreme gravity and test general relativity. The story extends OpenAI's "Year of Science" positioning and its push to embed Codex in scientific research workflows, following the DOE/national-lab collaborations and GPT-Rosalind life-sciences track.
Scoring impact: Modest positive for Distribution & Reach (cultivating the scientific-research user segment for Codex) and Research & Innovation narrative (frontier-science adoption). Low-signal marketing content rather than a product or capability launch.
Sources:
Published: 2026-06-10 | Logged: 2026-06-11T00:00Z | Area: Distribution & Reach, Business & Market, Compute & Infra
OpenAI and Oracle announced that Oracle Cloud Infrastructure (OCI) customers will be able to apply eligible Oracle Universal Credits toward OpenAI frontier models and Codex, rolling out in the coming weeks. The deal lets enterprises access OpenAI through their existing Oracle purchasing workflows and committed cloud spend, extending the multi-cloud procurement playbook (after AWS Bedrock GA) and aligning AI adoption with planned Oracle investments.
Scoring impact: Positive for Distribution & Reach (new enterprise procurement channel via Oracle's installed base) and Business & Market (frictionless monetization through committed-spend conversion). Reinforces Compute & Infra ties given OpenAI's large Oracle/Stargate capacity commitments.
Sources:
OpenAI submitted a draft registration statement to the SEC for a proposed initial public offering, following Anthropic's confidential filing roughly a week earlier. The company β last valued at $852 billion post-money β said it has not set timing and noted some priorities are easier to pursue while private. The filing sets up a major public-market test of the AI sector and signals OpenAI's intent to broaden its capital base for compute and infrastructure commitments.
Scoring impact: Positive for Business & Market (access to public capital, validation of scale) though governance and disclosure scrutiny rise.
Area: Safety & Alignment, Product & Platform
OpenAI introduced Lockdown Mode, an opt-in advanced security setting available to all logged-in ChatGPT users. When enabled, it restricts network-enabled capabilities β live web browsing, deep research, agent mode, file downloads, and some web-derived image support β to reduce the risk of data exfiltration from prompt-injection attacks. The feature targets enterprise and security-conscious users as agentic browsing expands the attack surface.
Scoring impact: Positive for Safety & Alignment (concrete anti-exfiltration control for agentic use) and supportive for Product & Platform (enterprise-grade security posture).
Area: Model Quality, Research & Innovation, Product & Platform
OpenAI released an updated GPT-Rosalind aimed at life-sciences research, combining GPT-5.5's agentic coding and tool-use with stronger domain intelligence in drug discovery, medicinal chemistry, and genomics. The update adds new plugins for evidence retrieval and bioinformatics workflows, extending OpenAI's vertical push into scientific R&D.
Scoring impact: Positive for Research & Innovation and Model Quality (domain-specialized frontier capability); incremental Product & Platform lift via a targeted scientific vertical.
Area: Product & Platform, Distribution & Reach
OpenAI rolled out an enterprise-focused Codex expansion this week, adding role-specific business plugins spanning sales, data analytics, creative production, product design, public-equity investing, and investment banking. It also previewed Sites β which turns plans, analyses, and ideas into shareable interactive apps and websites β plus annotations that let teams refine documents, spreadsheets, and slides in place.
Scoring impact: Positive for Product & Platform and Distribution & Reach, broadening Codex from a coding tool toward a general enterprise workspace agent.
Area: Research & Innovation, Product & Platform
OpenAI announced "Dreaming: Better memory for a more helpful ChatGPT," a research initiative aimed at strengthening ChatGPT's long-term memory and personalization. The disclosure accompanied a broader wave of product updates including real-time audio models for conversational agents, translation and transcription, a self-serve Ads Manager inside ChatGPT, and Codex Computer Use on Windows.
Scoring impact: Supports Research & Innovation and Product & Platform; the self-serve Ads Manager further signals a maturing ChatGPT ad business that could lift Business & Market over time.
Area: Distribution & Reach, Product & Platform, Business & Market
OpenAI frontier models and Codex moved to general availability on AWS through Amazon Bedrock, opening commercial and GovCloud regions to OpenAI capabilities under AWS-native security, governance, procurement, billing, and compliance controls. The launch operationalizes the previously announced OpenAI/AWS preview and gives millions of AWS customers a direct procurement path to OpenAI without leaving their existing cloud accounts.
Scoring impact: Major positive for Distribution & Reach (frontier models now first-class citizens on the largest cloud provider) and Business & Market (multi-cloud monetization reduces Azure dependence). Reinforces the same dynamic that drove the May 20 Bedrock preview into a much wider enterprise channel.
Sources:
Area: Product & Platform, Distribution & Reach, Business & Market
ChatGPT shipped a live job search experience and resume formatting tools on June 1. The feature surfaces personalized live listings and freelance opportunities pulled from Indeed, Upwork, Appcast, and other sources, lets users follow source-site links to apply, and provides upload/create/download flows for polished resumes. Job search is available to U.S. users on Free, Go, Plus, and Pro plans; resume formatting is available globally in English on the web for all plans.
Scoring impact: Positive for Product & Platform (extends ChatGPT into a high-intent vertical with proprietary content surfacing) and Distribution & Reach (expands the bridge between ChatGPT and external commerce/jobs partners). Modest Business & Market upside if engagement and partner referral economics meaningfully expand monetization beyond subscriptions.
Sources:
Area: Safety & Alignment, Business & Market
Florida Attorney General James Uthmeier filed a first-of-its-kind state lawsuit against OpenAI and CEO Sam Altman tied to violent incidents allegedly linked to ChatGPT, accusing the company of consumer-safety failures. The case lands as Anthropic files for its IPO and OpenAI itself prepares to file, intensifying regulatory and litigation overhang on the frontier-lab IPO cycle.
Scoring impact: Negative for Safety & Alignment (consumer-safety enforcement action) and Business & Market (regulatory/litigation drag during pre-IPO period); could pressure OpenAI to accelerate already-announced sensitive-conversation safeguards.
Sources:
Area: Product & Platform, Model Quality
OpenAI expanded Codex with Computer Use on Windows, letting eligible users see, click, and type in Windows apps while testing and refining builds; the update also adds remote continuation from mobile/Mac, faster browsing, new Codex Profiles with usage/token activity. Separately, GPT-5.5 Instant was updated with clearer, more natural responses and now supports writing and coding blocks directly in chat (Canvas is being removed from GPT-5.5 Instant and Thinking). OpenAI set retirement dates: o3 sunset from ChatGPT on Aug 26, 2026 (90-day) and GPT-4.5 on June 27, 2026 (30-day).
Scoring impact: Positive for Product & Platform (Windows agentic surface broadens Codex reach, Profiles deepen developer telemetry) and Model Quality narrative (Instant tier refresh). Modest cleanup of legacy model SKUs simplifies the lineup.
Area: Business & Market, Product & Platform
Multiple reports indicate OpenAI is testing upgraded ad formats inside ChatGPT, including larger images, customizable call-to-action buttons, and dedicated e-commerce layouts featuring pricing and customer reviews. The shift comes alongside research that 44% of enterprise e-commerce merchants are already integrating agentic commerce protocols, with another 32% expecting to do so within six months, and merchants anticipating AI shopping agents could account for up to 30% of transactions within three years.
Scoring impact: Positive Business & Market (early monetization beyond subscriptions/API; agentic commerce optionality) but risks Product & Platform trust given competitor positioning (Anthropic/Claude) as ad-free.
Area: Safety & Alignment, Product & Platform
OpenAI announced improvements to ChatGPT's behavior in sensitive conversations, including better recognition of emerging risk over time and new safety summaries that persist across chats. The updates help the model de-escalate, refuse harmful details, and respond more carefully in high-risk suicide, self-harm, and harm-to-others scenarios. The rollout coincides with broader Codex updates covering goals, permissions, plugins, extensions, and app-server workflows.
Scoring impact: Marginal positive for Safety & Alignment by demonstrating ongoing investment in real-world deployment safeguards; small Product & Platform boost from Codex polish.
Published: 2026-05-22 | Logged: 2026-05-23T08:00Z | Area: Business & Market
OpenAI confidentially filed its S-1 IPO prospectus with the SEC on Friday, May 22, working with Goldman Sachs and Morgan Stanley. The company is targeting a public listing as early as Q3-Q4 2026, with a valuation reportedly between $852 billion and $1 trillion. This makes the previously reported preparation phase official.
Scoring impact: Strongly positive for Business & Market β confirms move toward public capital markets at trillion-dollar valuation and unlocks new disclosure-driven capital structure.
Sources:
Published: 2026-05-22 | Logged: 2026-05-23T08:00Z | Area: Product & Platform, Distribution & Reach
OpenAI rolled out a preview of a new personal finance experience in ChatGPT for U.S. Pro users, allowing people to securely connect bank/investment accounts, view a money dashboard, and ask context-grounded financial questions on web and iOS. Codex also gained Mac Appshots, GA Goal mode, in-app browser annotations, and locked computer use on Mac.
Scoring impact: Positive for Product & Platform (deeper consumer financial integration) and Distribution & Reach (expands ChatGPT's role as personal AI hub).
Sources:
Published: 2026-05-20 | Logged: 2026-05-21T09:00Z | Area: Business & Market
OpenAI is preparing to file confidentially for an IPO in the coming weeks, possibly as soon as Friday, with Goldman Sachs and Morgan Stanley as lead underwriters. The company is targeting a public debut in the fall (possibly September) and has committed to spending more than $1.4 trillion on physical infrastructure in the coming years. The filing follows Elon Musk losing his lawsuit threatening OpenAI's structure earlier in the week.
Scoring impact: Major positive for Business & Market (path to substantial public-market capital). Indirectly supports Compute & Infra given declared $1.4T infrastructure commitments.
Sources:
Published: 2026-05-20 | Logged: 2026-05-21T09:00Z | Area: Distribution & Reach, Product & Platform, Compute & Infra
AWS announced an expanded OpenAI partnership: GPT-5.5 and GPT-5.4 are coming to Amazon Bedrock in limited preview, alongside Codex (via CLI, desktop app, and VS Code extension). AWS also added OpenAI-compatible API support to Amazon SageMaker endpoints, and launched Bedrock Managed Agents powered by the OpenAI harness. The moves significantly widen the distribution surface for OpenAI frontier models inside enterprise AWS estates.
Scoring impact: Strong positive for Distribution & Reach (Bedrock = primary enterprise AI surface). Also supportive for Product & Platform (Codex on Bedrock) and Compute & Infra (multi-cloud delivery).
Sources:
Published: 2026-05-20 | Logged: 2026-05-24T09:00Z | Area: Model Quality, Research & Innovation
OpenAI announced that one of its internal general-purpose reasoning models autonomously disproved Paul ErdΕs's 1946 unit-distance conjecture in discrete geometry, producing a 125-page proof from the problem statement alone β no problem-specific training, no retrieval, no step-by-step human guidance. The construction surprises the field by reaching from elementary geometry into algebraic number theory (Golod-Shafarevich class field towers), establishing an infinite family of planar configurations that beat the square grid and scale as n^(1+Ξ΄) with Ξ΄ β₯ 0.014, per Princeton's Will Sawin in a companion paper. Fields medalist Tim Gowers called the result "a milestone in AI mathematics," and Princeton combinatorialist Noga Alon called it "an outstanding achievement." Coverage widened materially May 21-23.
Scoring impact: Major Research & Innovation boost β first reported instance of an AI autonomously solving a prominent open problem central to a field of mathematics rather than a competition benchmark. Reinforces Model Quality positioning for OpenAI's unreleased reasoning frontier. Strengthens Talent & Org narrative for recruiting frontier scientific researchers.
Sources:
Published: 2026-05-19 | Logged: 2026-05-20T09:00Z | Area: Distribution & Reach, Business & Market, Product & Platform
OpenAI and Dell Technologies announced a partnership to bring Codex into hybrid and on-premises enterprise environments via the Dell AI Data Platform and Dell AI Factory. The deal targets enterprises (UK financial services, healthcare, government) that cannot send data to public cloud, letting Codex, ChatGPT Enterprise, and other API-based solutions interface with Dell-managed data and infrastructure. OpenAI disclosed Codex now has more than 4 million weekly developers and is expanding beyond coding into agentic workflows including code review, test coverage, incident response, lead qualification, follow-up writing, and cross-system coordination. This is OpenAI's first explicit hybrid/on-prem enterprise distribution play.
Scoring impact: Strong Distribution & Reach (on-prem channel via Dell's massive enterprise footprint). Positive Business & Market (unlocks data-sovereign verticals that were previously gated). Modest Product & Platform lift from deeper enterprise-data integration.
Sources:
Published: 2026-05-18 | Logged: 2026-05-19T09:00Z | Area: Distribution & Reach, Business & Market, Talent & Org
OpenAI and the Government of Malta announced a world-first national partnership: every Maltese resident (~574,250 people) can earn one year of free ChatGPT Plus after completing an AI literacy course developed by the University of Malta. The first phase launches in May, with Malta's Digital Innovation Authority handling distribution. The deal is the latest under OpenAI's "OpenAI for Countries" initiative, aimed at converting national governments into distribution partners and surfacing population-scale usage data and AI-skills infrastructure.
Scoring impact: Positive for Distribution & Reach (entire country onboarded as a Plus market through a state channel). Modest Business & Market lift via the OpenAI for Countries playbook and a likely template for other small/mid-size nations. Light Talent & Org positive through formal national AI-literacy curriculum tie-in.
Sources:
Published: 2026-05-15 | Logged: 2026-05-18T09:00Z | Area: Product & Platform, Talent & Org, Business & Market
OpenAI told employees Friday (May 15) that co-founder and president Greg Brockman will permanently lead all product strategy, merging ChatGPT, Codex, and the developer API into one organization. Thibault Sottiaux β the engineer who built Codex into one of OpenAI's fastest-growing products β now leads the combined core product and platform across consumer, enterprise, and developer surfaces. Nick Turley, who oversaw ChatGPT's expansion since 2022, moves into a role focused on enterprise products and critical industries. The reorganization formalizes the previously reported desktop "super app" project bringing ChatGPT, Codex, and the Atlas browser into a single agentic application. The news lands four days before Google I/O 2026.
Scoring impact: Positive for Product & Platform (unified surface aligns with the agentic super-app strategy). Talent & Org signal as product leadership consolidates under Brockman/Sottiaux. Business & Market relevant given timing ahead of IPO and competitive positioning.
Sources:
Published: 2026-05-15 | Logged: 2026-05-15T09:00Z | Area: Business & Market, Compute & Infra
CFO Sarah Friar said OpenAI may raise additional capital after closing what she described as the largest private fundraising round ever (a recent $122B round), citing surging demand and the persistent gap between the compute OpenAI needs and what it can currently afford. Friar said the round provides "a lot of optionality," with further raises dependent on demand, revenue growth, cash flow, and capacity buildout.
Scoring impact: Mixed Business & Market β large capital base is positive, but ongoing dilution and capex pressure underscore the structural cost of leadership. Reinforces the Compute & Infra constraint as the binding scaling factor.
Sources:
Published: 2026-05-15 | Logged: 2026-05-16T09:00Z | Area: Product & Platform, Distribution & Reach, Data & Moats
OpenAI launched a personal finance toolset in preview for ChatGPT Pro subscribers in the U.S., partnered with Plaid to let users connect accounts at over 12,000 institutions including Schwab, Fidelity, Chase, Robinhood, American Express, and Capital One. Users can ask ChatGPT to analyze spending, plan savings, and provide tailored financial guidance. The launch extends ChatGPT from a generalist assistant into a vertical (consumer finance) historically dominated by fintech apps and big-bank advisory services.
Scoring impact: Strong positive for Product & Platform (new high-value vertical surface) and Distribution & Reach (Plaid plugs ChatGPT into household financial accounts). Modest Data & Moats lift as opt-in financial signals become a proprietary personalization input.
Sources:
Published: 2026-05-14 | Logged: 2026-05-15T09:00Z | Area: Distribution & Reach, Business & Market
Accenture Federal Services and OpenAI announced a strategic collaboration to help U.S. federal agencies adopt, migrate, and scale advanced AI. AFS will enable all 15,000 of its professionals with secure, governed access to OpenAI's latest models, paired with training, role-based access controls, and continuous oversight, to accelerate AI adoption in agencies including via the GenAI.mil platform.
Scoring impact: Positive Distribution & Reach (deep federal-agency systems-integrator channel) and Business & Market (expanded public-sector revenue surface alongside the May 1 Pentagon agreement).
Sources:
Published: 2026-05-14 | Logged: 2026-05-15T09:00Z | Area: Safety & Alignment, Business & Market
OpenAI publicly endorsed the creation of a U.S.-led global AI governance body that would include China as a member, hours before a TrumpβXi meeting. The position is a notable shift in tone for a U.S. frontier lab and continues OpenAI's pattern of staking out policy positions on dual-use capabilities (red lines for military use, Trusted Access for cyber, AISI evaluations).
Scoring impact: Modest positive for Safety & Alignment (public commitment to international governance). Business & Market impact depends on whether such a body materializes and how it shapes export controls.
Sources:
Published: 2026-05-14 | Logged: 2026-05-15T09:00Z | Area: Distribution & Reach, Business & Market
Bloomberg reported the two-year-old AppleβOpenAI partnership has become strained, with OpenAI saying it hasn't seen the expected benefits from the deal and reportedly preparing possible legal action. The friction comes as Apple is widely expected to lean on Google's Gemini to power its delayed Siri overhaul ahead of WWDC 2026.
Scoring impact: Negative Distribution & Reach (risk of losing the default iPhone AI surface) and Business & Market (loss of a flagship consumer integration if the relationship deteriorates further).
Sources:
Published: 2026-05-14 | Logged: 2026-05-15T09:00Z | Area: Safety & Alignment, Product & Platform
OpenAI said it found no evidence of customer data breaches or system compromise following a supply-chain attack on the TanStack open-source npm library, which is used in some OpenAI tooling. The disclosure addresses concerns from developers and enterprise customers and emphasizes OpenAI's containment posture for third-party supply-chain incidents.
Scoring impact: Mildly positive for Safety & Alignment (clear, fast incident response and customer communication). Neutral for Product & Platform.
Sources:
Published: 2026-05-13 | Logged: 2026-05-14T09:00Z | Area: Model Quality, Product & Platform
OpenAI released three new Realtime voice models in the API: GPT-Realtime-2 (its first voice model with GPT-5-class reasoning), GPT-Realtime-Translate (real-time multilingual speech-to-speech), and GPT-Realtime-Whisper (streaming transcription). The models target live voice agents and translation use cases, complementing the prior week's voice intelligence model launches.
Scoring impact: Positive for Product & Platform (broader voice/audio API portfolio). Modest Model Quality boost (GPT-5-class reasoning extended into the realtime voice stack).
Sources:
Area: Business & Market, Distribution & Reach, Talent & Org
OpenAI formally launched the OpenAI Deployment Company on May 11, backed by more than $4 billion in initial capital from a 19-firm consortium led by TPG with co-lead founding partners Advent, Bain Capital, and Brookfield. Simultaneously, OpenAI agreed to acquire Tomoro, a Scotland-based applied AI consulting firm formed in 2023 with clients including Mattel, Red Bull, Tesco, and Virgin Atlantic, adding roughly 150 Forward Deployed Engineers and Deployment Specialists from day one. CRO Denise Dresser said enterprises are struggling to keep up with "compounded innovation" in AI models, and the venture is designed to embed engineers directly inside client organizations. The structure remains majority-owned and controlled by OpenAI.
Scoring impact: Reinforces Business & Market (enterprise services revenue stream now operationalized) and Distribution & Reach (deeper client embed via Tomoro's customer base). Positive Talent & Org (150 forward-deployed engineers added).
Sources:
Area: Model Quality, Product & Platform, Research & Innovation
OpenAI introduced three new voice models in the API: GPT-Realtime-2, its first voice model with GPT-5-class reasoning for natural multi-turn dialog; GPT-Realtime-Translate, which streams speech translation from 70+ input languages into 13 output languages while keeping pace with the speaker; and GPT-Realtime-Whisper, a new streaming speech-to-text that transcribes live as the speaker talks. The trio is positioned to let developers build talk/transcribe/translate apps without bolting together separate vendors.
Scoring impact: Strengthens Model Quality (frontier voice reasoning) and Product & Platform (more capable voice API). Helps Distribution & Reach via lower-friction voice integrations for developers.
Area: Safety & Alignment, Product & Platform
OpenAI began rolling out Trusted Contact, an opt-in setting that lets a ChatGPT user nominate a person to be notified if automated systems and trained reviewers detect a serious suicide-related safety signal in conversation. The contact receives an invitation to participate; once they accept, ChatGPT may reach out to them to check in on the user during a flagged event. This pairs with broader safety routing changes and parental controls OpenAI has shipped over the last few months.
Scoring impact: Positive for Safety & Alignment (concrete in-product safety mechanism for high-risk conversations); modest plus for Product & Platform.
Published: 2026-05-05 | Logged: 2026-05-06T09:00Z | Area: Model Quality, Product & Platform, Safety & Alignment
OpenAI rolled out GPT-5.5 Instant to all ChatGPT users, replacing GPT-5.3 Instant as the default model and shipping in the API as chat-latest. OpenAI says GPT-5.5 Instant produces 52.5% fewer hallucinated claims than GPT-5.3 Instant on high-stakes prompts in medicine, law, and finance, with tighter, less verbose responses and fewer "gratuitous emojis." The model can use its search tool to refer back to past conversations, files, and Gmail for more personalized answers (Plus and Pro users on web), and ChatGPT now surfaces memory sources across all models so users can delete or correct what's used. OpenAI also published an openai/gpt-5-5-instant model card.
Scoring impact: Positive for Model Quality (large hallucination reduction on high-stakes tasks). Strong Product & Platform improvement (default model upgrade for the entire ChatGPT user base). Safety & Alignment boost from transparent memory sourcing and improved factuality.
Sources:
Published: 2026-05-04 | Logged: 2026-05-05T09:00Z | Area: Business & Market, Distribution & Reach
OpenAI finalized a $10 billion joint venture called "The Deployment Company" anchored by TPG, with 19 investors including Brookfield Asset Management, Advent, Bain Capital, and SoftBank. The structure includes ~$4B+ from outside investors and approximately $1.5B of OpenAI's own capital, with a 17.5% guaranteed annual return over five years for partners. The venture is designed to help mid-size and large enterprises bring OpenAI models into core operations, mirroring Anthropic's separately announced $1.5B Wall Street JV the same day. Both deals put the model labs in direct competition with the global consulting industry for AI transformation work.
Scoring impact: Major boost to Business & Market (new enterprise services revenue stream, significant outside capital). Strong Distribution & Reach (PE-owned portfolio companies become a deployment channel).
Sources:
Published: 2026-05-01 | Logged: 2026-05-03T18:00Z | Area: Distribution & Reach, Business & Market, Safety & Alignment
OpenAI is among seven companies (alongside SpaceX, Google, Nvidia, Reflection, Microsoft, and AWS) that signed agreements with the Department of Defense to deploy AI tools in the Pentagon's Impact Level 6 and IL7 classified networks via the GenAI.mil platform. OpenAI publicly outlined three "red lines" guiding the work: no mass domestic surveillance, no use to direct autonomous weapons systems, and no high-stakes automated decisions. The deal explicitly excludes Anthropic, which the Pentagon labeled a supply-chain risk after the company refused unrestricted military access.
Scoring impact: Positive for Distribution & Reach and Business & Market (defense channel). Mixed for Safety & Alignment (structured access with stated red lines, but expanding into sensitive military use cases).
Sources:
Published: 2026-05-01 | Logged: 2026-05-03T18:00Z | Area: Model Quality, Safety & Alignment, Product & Platform
OpenAI began rolling out GPT-5.5-Cyber to a hand-picked circle of "cyber defenders" through its Trusted Access for Cyber program, requiring Advanced Account Security from June 1 for individual members. The UK's AI Security Institute called it "one of the strongest models we have tested on our cyber tasks" and noted it is only the second system ever to complete one of its multi-step attack simulations end-to-end. Tiered, gated access continues OpenAI's pattern (after GPT-5.4-Cyber) of releasing dual-use capabilities behind structured-access controls.
Scoring impact: Boost to Model Quality (frontier-level offensive cyber capability validated by AISI). Positive for Safety & Alignment (structured access). Adds specialized Product & Platform variant.
Sources:
Published: 2026-04-30 | Logged: 2026-05-02T09:00Z | Area: Product & Platform, Safety & Alignment
ChatGPT rolled out Advanced Account Security for personal accounts, adding stronger sign-in options, stricter recovery, login notifications, and session controls to help protect against unauthorized access and account takeover.
Scoring impact: Minor positive for Product & Platform and Safety & Alignment (improved user security).
Sources:
Published: 2026-04-28 | Logged: 2026-05-02T09:00Z | Area: Distribution & Reach, Business & Market
OpenAI launched GPT-5.5 and other frontier models on Amazon Bedrock, brought Codex to AWS, and added Bedrock Managed Agents powered by OpenAI. This marks a significant loosening of the previous Microsoft-linked exclusivity structure, opening distribution through Amazon and potentially Google.
Scoring impact: Significantly raises Distribution & Reach (multi-cloud availability). Positive for Business & Market as enterprise access widens beyond Azure.
Sources:
Published: 2026-04-23 | Logged: 2026-05-02T09:00Z | Area: Model Quality, Product & Platform
OpenAI released GPT-5.5, described as its smartest and most intuitive model yet, with faster agentic coding, stronger computer use, and better knowledge work in ChatGPT and Codex. GPT-5.5 Thinking is available in ChatGPT for eligible paid plans, while GPT-5.5 Pro is available to Pro, Business, Enterprise, and Edu plans. OpenAI surpassed $25B in annualized revenue.
Scoring impact: Major boost to Model Quality (new frontier model). Strengthens Product & Platform with deeper ChatGPT/Codex integration.
Sources:
Published: 2026-04-16 | Logged: 2026-05-02T09:00Z | Area: Model Quality, Research & Innovation
OpenAI introduced GPT-Rosalind, a frontier reasoning model for biology, drug discovery, and translational medicine. Available as research preview in ChatGPT, Codex, and API for qualified customers. Working with Amgen, Moderna, Allen Institute, Thermo Fisher Scientific. Named after Rosalind Franklin.
Scoring impact: Expands Model Quality into specialized vertical (life sciences). Positive for Research & Innovation (domain-specific frontier model).
Sources:
Published: 2026-04-14 | Logged: 2026-05-02T09:00Z | Area: Product & Platform, Safety & Alignment
OpenAI released GPT-5.4-Cyber, designed for defensive cybersecurity tasks with more permissive access for vetted users through a tiered cyber access program.
Scoring impact: Positive for Product & Platform (specialized model variants). Positive for Safety & Alignment (structured access for sensitive capabilities).
Sources: