OpenAI launched GPT-5.5 (and GPT-5.5 Pro) on 23 April, describing it as its smartest, most intuitive and most agentic model yet. President Greg Brockman framed the release as a meaningful step toward an OpenAI super app that combines reasoning, computer use, coding and document creation in a single experience. It lands less than two months after GPT-5.4, signalling an aggressive cadence as competitive pressure builds from Anthropic and DeepSeek.
The headline upgrades include stronger agentic coding, computer use, data analysis and end-to-end document creation; a 1M-token context window in the API matching the largest commercial frontier models; improved performance on mathematics, scientific research and long-horizon reasoning; and rollout to Plus, Pro, Business and Enterprise users in ChatGPT and Codex first, with API availability following on 24 April.
The strategic context is what makes this release matter beyond the benchmarks. GPT-5.5 sits at the centre of OpenAI’s race to anchor enterprise spend before Anthropic’s IPO and Google’s Gemini Enterprise push close the gap. The model launched alongside Workspace Agents, an expanded Codex partner network with the world’s largest consultancies, and the new union-built Stargate Michigan data centre. Codex’s weekly developer count grew from three million to four million in the two weeks before launch, an acceleration OpenAI is clearly racing to capitalise on.
The cadence between GPT-5.x releases is collapsing into something close to continuous deployment. Two months between GPT-5.4 and GPT-5.5 is a fundamentally different rhythm to the one-major-release-a-year pattern that defined AI through 2024. Combined with the Codex rebuild last week and Workspace Agents this week, OpenAI is shipping at the pace of a SaaS platform, not a research lab. Nvidia rolled GPT-5.5 out organisation-wide on day one, and within hours OpenAI was managing minor incidents flagging elevated errors on the new model in Codex. That speed of adoption is itself the story: enterprises now expect to deploy frontier-model upgrades the day they ship.
For enterprises, the practical implication is that “AI strategy” needs to absorb a continuous-upgrade reality. The model your team trained against last quarter is already a release behind. For OpenAI, GPT-5.5 needs to lift enterprise ARR fast enough to justify the $122B funding round and pull ChatGPT users away from increasingly capable rivals. Google announced a $40B Anthropic commitment the day after GPT-5.5 shipped, which tells you exactly how seriously the rest of the market is taking this race.
The bottom line: super-app strategy is no longer a slide deck phrase. It is the lens through which to read every OpenAI move this year, and the test it has to pass is whether end users start treating ChatGPT as the desktop, not the chatbot.
The Foundations of a Generational AI Economy Are Being Poured. This week looked less like a flurry of product announcements and more like the long-term structures of the AI era being locked into place. Google’s $40B Anthropic deal, Microsoft’s $18B Australia commitment, SpaceX’s $60B Cursor option, Google Cloud’s $750M agentic-AI partner fund, OpenAI’s union-built Stargate Michigan data centre and DeepSeek’s open-source 1M-context preview all push in the same direction: power, capital and labour are being committed to multi-year structures around frontier AI.
Layoffs at Meta and Microsoft are the other side of the same coin. The story of 2026 is no longer “who has the best model” but “who controls the chips, the customers and the workforce.” The winners of the next two years will lock in compute, consultancy capacity and skilled labour now, while everyone else is still optimising prompts.
In a single seven-day window, OpenAI shipped ChatGPT Images 2.0, Kuaishou rolled out a native 4K mode for Kling 3.0, Odyssey ML released Odyssey-2 Max (a 3x larger world model that beats Sora and Veo on physics scores), and OpenAI confirmed the consumer Sora app will be discontinued on 26 April. The visual-AI market is consolidating fast around a small number of credible frontier players and three distinct lanes: thinking-driven stills, native 4K cinematic video, and continuous interactive simulation.
The strategic read is that OpenAI is using gpt-image-2 to reassert dominance in stills while quietly rebuilding its video strategy around enterprise and Hollywood partners rather than a consumer app. Kling and Odyssey are no longer fringe alternatives. Expect Adobe, Canva and Figma to embed gpt-image-2 inside their creative suites within the quarter, and watch closely for the first credible “thinking video” release. That is the next obvious shoe to drop.
Read Odyssey-2 Max →