Is ChatGPT Down?

PRODUCT (2026-05-14, TODAY): OpenAI launched **Codex on mobile** -- the Codex coding agent is now integrated into the ChatGPT mobile app (iOS + Android), preview, all plans. Users can monitor, approve, and manage Codex coding sessions from phone while files / credentials / local setup stay on the host machine. Concurrent launches: **Remote SSH** (Codex can SSH into your machines from the cloud), **Hooks** (programmable callbacks at each step of Codex execution), **programmatic access tokens** for business / enterprise, and **HIPAA-compliant Codex** on ChatGPT Enterprise. OpenAI disclosed **4M+ weekly Codex users**. Vendor primary: openai.com/index/work-with-codex-from-anywhere/. Material for the broader 'where do you sit when you're driving an AI coding agent' question -- Claude Code shipped Remote Agents at the 5/6 Code with Claude keynote (laptop control from phone); OpenAI shipped Codex mobile + Remote SSH today; Cursor shipped MS Teams integration 5/11 + cloud Dev Environments 5/13. The 'agent goes where you go' battle is escalating fast

API DEPRECATION (2026-05-08): OpenAI added a new entry to the deprecation page -- **gpt-5.2-chat-latest** and **gpt-5.3-chat-latest** model snapshots are scheduled to retire. These are distinct from the 4/22 legacy snapshot batch (gpt-5-chat-latest + gpt-5.1-codex retire 7/23; o1-2024-12-17 + ft-gpt-3.5-turbo + ft-gpt-4 retire 10/23) -- the 5/8 entry covers the GPT-5.2 / GPT-5.3 'chat-latest' rolling-pointer snapshots specifically. If you have code or fine-tunes pinned to either snapshot ID, schedule migration. Replacement guidance per the deprecation page: move to the current 'chat-latest' alias (which routes to GPT-5.5) or pin a versioned snapshot ID for stability

CORPORATE (2026-05-11): OpenAI launched 'The Deployment Company' -- a new majority-owned OpenAI subsidiary with $4B initial investment from 19 partners (TPG, Goldman Sachs, Bain Capital, McKinsey + 15 others). Mission: embed Forward Deployed Engineers (FDEs) inside client organizations to accelerate enterprise AI deployment beyond what an API + consulting can deliver. Concurrent: acquired Tomoro (~150 AI engineers) to seed the FDE bench. Strategic frame: a delivery layer for GPT-5.x / Codex / future models inside Fortune-500 client orgs, competing with Accenture / McKinsey on AI implementation but with vendor-aligned incentives. NOT a new ChatGPT feature -- this is a services-arm spin-out. Implications for buyers: lock-in risk if you sign with the Deployment Company (deploys against OpenAI models specifically); upside is faster time-to-value on Codex / GPT-5.5 agentic workflows. openai.com/index/openai-launches-the-deployment-company/ returned 403 on direct WebFetch (anti-bot, expected); confirmed via OpenAI's official X announcement + tier-1 press (Axios, PYMNTS, CNBC).

GPT-5.5 'Spud' launched 2026-04-23 in ChatGPT (Plus/Pro/Business/Enterprise) + Codex (Plus/Pro/Business/Enterprise/Edu/Go). Model IDs gpt-5.5 and gpt-5.5-pro. ChatGPT Free tier NOT included. OpenAI-published launch benchmarks: Terminal-Bench 2.0 82.7% (SOTA, +7.6 over GPT-5.4), SWE-Bench Pro 58.6%, GDPval 84.9% (SOTA), OSWorld-Verified 78.7%, ARC-AGI-2 85.0% (+11.7 over GPT-5.4's 73.3 on ARC-AGI-1), FrontierMath Tier 4 35.4% base / 39.6% Pro, Humanity's Last Exam 52.2% (with tools), MRCR 512K-1M 74.0% (+37.4 over GPT-5.4's 36.6 -- the biggest material delta). Cyber capability classified 'High' under OpenAI's Preparedness Framework (below Critical). No GPT-5.4 deprecation announced -- GPT-5.4 remains live. Third-party verification of launch benchmarks pending

GPT-5.5 API SHIPPED 2026-04-24 (~24 hours after ChatGPT consumer launch). Both gpt-5.5 ($5/$30 per 1M tokens) and gpt-5.5-pro ($30/$180 per 1M) are now live via OpenAI API. Cached-input pricing revealed: $0.50/M cached input (90% discount vs standard input), Batch + Flex tiers at 50% of standard, Priority at 2.5x. Faster API rollout than expected -- yesterday's launch post said 'coming very soon' and it shipped in under 36 hours

GPT-5.5 pricing jumped 2x vs GPT-5.4 on the API: $5/$30 per 1M tokens input/output (vs $2.50/$15 for GPT-5.4). GPT-5.5 Pro is $30/$180 per 1M -- the most expensive frontier LLM API on the market today. Batch + Flex = 50% of standard rate; Priority processing = 2.5x standard. For budget-sensitive workloads, GPT-5.4 at half the token cost remains a defensible choice where GPT-5.5's agentic/coding gains don't apply