OpenAI releases "Spud" GPT-5.5 model

```json { "title": "OpenAI Releases GPT-5.5 'Spud': What You Need to Know", "metaDescription": "OpenAI launched GPT-5.5, codenamed 'Spud,' on April 23, 2026. Here's what's new, who can access it, and why it matters for AI productivity.", "content": "<h2>OpenAI Releases GPT-5.5 'Spud,' Its Most Capable Model Yet</h2><p>OpenAI released GPT-5.5, internally codenamed \"Spud,\" on Thursday, April 23, 2026, marking what the company describes as its most capable AI model to date and its first fully retrained base model since GPT-4.5. The release came less than two months after OpenAI launched GPT-5.4 on March 5, 2026, and just one week after competitor Anthropic unveiled its latest model, Claude Mythos Preview — underscoring the increasingly compressed timelines defining the frontier AI race.</p><p>GPT-5.5 is rolling out immediately to paid ChatGPT and Codex subscribers across Plus, Pro, Business, and Enterprise tiers. API access has been delayed pending additional cybersecurity safeguards. OpenAI is also releasing GPT-5.5 Pro, a higher-accuracy variant available to Pro, Business, and Enterprise users.</p><h2>What GPT-5.5 Can Do: Agentic AI and Autonomous Workflows</h2><p>GPT-5.5 is engineered around a core idea: doing more with less human direction. The model is designed to handle complex, multi-step tasks — writing and debugging code, researching online, analyzing data, operating software, and producing documents — with minimal guidance from the user. OpenAI frames this as a significant step toward genuinely agentic computing, where AI doesn't just respond to prompts but anticipates what needs to happen next.</p><p>Greg Brockman, President and co-founder of OpenAI, described the shift in practical terms: <em>"What is really special about this model is how much more it can do with less guidance. It can look at an unclear problem and figure out just what needs to happen next. It really, to me, feels like it's setting the foundation for how we're going to use computers, how we're going to do computer work going forward."</em></p><p>On raw performance, OpenAI's own benchmarking data — as reported by Inc. — shows GPT-5.5 outperformed or tied with human workers on approximately 85% of benchmarked tasks, compared to 83% for GPT-5.4 and 80% for Anthropic's Opus 4.7. On Terminal-Bench 2.0, a benchmark testing complex command-line workflows requiring planning, iteration, and tool coordination, GPT-5.5 reaches 82.7%. On Artificial Analysis's Coding Index, OpenAI says GPT-5.5 delivers state-of-the-art coding performance at roughly half the cost of competing frontier coding models.</p><p>Teams given early access to the model reported saving up to 10 hours of work per week, according to OpenAI.</p><h2>GPT-5.5 and the Path to an AI 'Super App'</h2><p>Brockman positioned GPT-5.5 as more than a standalone model upgrade. In briefings, he described the release as an additional step toward a planned \"super app\" that would combine ChatGPT, Codex, and an AI browser into a single unified platform for enterprise customers. Three million developers currently use Codex weekly as of April 2026, providing a significant existing user base for that vision.</p><p>Brockman characterized the broader shift in blunt terms: <em>"We are moving to a compute-powered economy."</em></p><p>GPT-5.5 was trained on Nvidia GPUs, consistent with past OpenAI models. Nvidia reports that its newest chips can slash the per-token cost of running advanced AI models like GPT-5.5 by up to 35x — a development that, if it holds at scale, would significantly reduce the economic barrier to enterprise AI deployment. Nvidia vice president of enterprise computing Justin Boitano told Axios that the model can act as a \"chief of staff,\" helping power agents already functioning as employees at Nvidia.</p><h2>Pricing, Access, and Cybersecurity Classification</h2><p>API pricing for GPT-5.5 is set at $5 per million input tokens and $30 per million output tokens — exactly double the rates OpenAI charged for GPT-5.4, which was priced at $2.50 per million input tokens and $15 per million output tokens. The price increase reflects both the model's increased capability and OpenAI's positioning of GPT-5.5 as a premium enterprise tool.</p><p>On cybersecurity, OpenAI was direct about the model's risk profile. The company assessed that GPT-5.5 meets criteria for a \"High\" cybersecurity risk classification — meaning it could amplify existing threats if misused — but does not exceed its internal \"Critical\" threshold. Mia Glaese, Vice President of Research at OpenAI, addressed the safeguard process directly: <em>"GPT-5.5 underwent extensive third-party safeguard testing and red teaming for cyber and bio [risks], and we've been iterating on our cyber safeguards for months with increasingly cyber capable models."</em></p><p>API access remains delayed while additional cybersecurity safeguards are finalized, though no specific timeline was given for when the API will become broadly available.</p><h2>Context: A Code Red Competitive Landscape</h2><p>The release of GPT-5.5 comes against a backdrop of sustained competitive pressure. OpenAI has reportedly been in an internal \"Code Red\" posture since at least December 2025, driven by significant gains from both Anthropic and Google in the enterprise AI market. Anthropic's annualized recurring revenue grew from $9 billion to $30 billion in the period leading up to GPT-5.5's launch — a trajectory that intensified urgency inside OpenAI ahead of the release.</p><p>Anthropic's Claude Mythos Preview, released approximately one week before GPT-5.5, attracted enterprise and Wall Street attention and set the immediate competitive backdrop for OpenAI's announcement. The pace of releases from both companies — measured now in weeks rather than quarters — signals a structural acceleration in the frontier AI development cycle that shows no sign of slowing.</p><p>Jakub Pachocki, Chief Scientist at OpenAI, framed the performance trajectory in optimistic but carefully bounded terms: <em>"We see pretty significant improvements in the short term, extremely significant improvements in the medium term."</em></p><p>Mark Chen, OpenAI's chief research officer, noted that GPT-5.5 \"shows meaningful gains on scientific and technical research workflows\" and suggested the model could have applications in drug discovery, though specific claims in that area remain to be validated through independent research.</p><h2>Early Enterprise Reactions</h2><p>The Bank of New York was among the organizations testing GPT-5.5 in the weeks before its public release. Leigh-Ann Russell, CIO of the Bank of New York, spoke to what the model delivered for a regulated financial institution: <em>"What we're actually seeing from 5.5, that I think is really important for a highly regulated institution, is the response quality — but also a really impressive hallucination resistance."</em></p><p>Hallucination resistance — the degree to which an AI model avoids generating plausible-sounding but factually incorrect outputs — is a particularly high-stakes concern in financial services, healthcare, and other regulated industries. Russell's comment suggests that for enterprise customers operating under compliance constraints, accuracy and reliability may matter as much as raw capability benchmarks.</p><h2>What's Next</h2><p>OpenAI has not announced a specific timeline for broad API availability of GPT-5.5, citing ongoing cybersecurity safeguard work as the determining factor. The company's broader roadmap, as outlined by Brockman, points toward the eventual integration of ChatGPT, Codex, and an AI browser into a unified enterprise platform — though no launch date for that combined product has been confirmed.</p><p>Pretraining for GPT-5.5 was completed on March 24, 2026, as confirmed by Sam Altman, meaning the gap between training completion and public release was approximately one month. Whether that cadence accelerates or stabilizes will likely depend on how quickly OpenAI can iterate on its safety and cybersecurity evaluation processes for each successive model.</p><p>The broader competitive dynamic between OpenAI, Anthropic, and Google in the enterprise AI market remains fluid. GPT-5.5's benchmark numbers represent a lead at the moment of release, but the pace of releases from all three companies means that lead is measured in weeks, not quarters.</p><p>For more tech news, visit our <a href=\"/news\">news section</a>.</p><h2>Why This Matters for Your Health and Productivity</h2><p>The shift toward agentic AI — systems that can interpret ambiguous instructions, plan multi-step tasks, and execute work autonomously — has direct implications for how individuals manage cognitive load, time, and energy. Early users of GPT-5.5 reported saving up to 10 hours of work per week, time that could be redirected toward higher-value thinking, recovery, or personal priorities. As AI tools become more capable of handling routine and complex cognitive tasks alike, understanding how to integrate them effectively into your workflow is increasingly a competitive advantage — not just a tech curiosity. At Moccet, we track these developments because the intersection of AI capability and human performance is central to what we do. <a href=\"/#waitlist\">Join the Moccet waitlist to stay ahead of the curve.</a></p>", "excerpt": "OpenAI released GPT-5.5, codenamed 'Spud,' on April 23, 2026 — its most capable model to date and first fully retrained base model since GPT-4.5. The launch came less than two months after GPT-5.4 and one week after Anthropic's Claude Mythos Preview, reflecting a rapidly accelerating AI development cycle. GPT-5.5 is now available to paid ChatGPT and Codex subscribers, with API access pending additional cybersecurity safeguards.", "keywords": ["GPT-5.5", "OpenAI Spud model", "agentic AI", "AI productivity", "frontier AI 2026"], "slug": "openai-releases-gpt-5-5-spud-model" } ```