The Dawn of Natively Omnimodal Intelligence
On April 23, 2026, OpenAI officially unveiled GPT-5.5, codenamed "Spud." While the industry was expecting an incremental update, Spud represents the first fully retrained base model since the GPT-4 era, built from the ground up to be natively omnimodal.
What is Native Omnimodality? Unlike previous models that used "adapters" or "vision encoders" to process images and audio, GPT-5.5 uses a single transformer architecture that treats text, pixels, and waveforms as equal tokens. This allows for zero-latency switching between modes.
The most significant leap in GPT-5.5 is its **Agentic Core**. OpenAI has shifted focus from "talking to AI" to "AI using computers." Spud is optimized for multi-step reasoning and tool orchestration, achieving record-breaking scores on the Agentic Reasoning Benchmark (ARB).
Key Features of GPT-5.5 "Spud"
1. Advanced "Computer Use" Mode
Spud can navigate standard desktop environments with human-like precision. It doesn't just see a screenshot; it understands the underlying UI hierarchy, allowing it to execute complex tasks like "organizing last month's invoices from my email into the accounting software and flagging discrepancies."
2. 10M Token Context Window
With a context window of 10 million tokens, GPT-5.5 can ingest entire codebases, legal libraries, or dozens of high-resolution video files simultaneously. This eliminates the need for complex RAG (Retrieval-Augmented Generation) for many medium-scale enterprise tasks.
3. Real-Time Physical Reasoning
Thanks to its native video processing, Spud can reason about physical space in real-time. This has immediate applications in robotics and AR/VR, where the model can guide users through physical repairs or assembly tasks by watching through a camera lens.
Why the name "Spud"?
While OpenAI hasn't officially commented on the codename, internal leaks suggest it stands for **"Systemic Processing & Unified Distribution."** Others jokingly suggest it’s a nod to its "low-power, high-yield" efficiency—running much faster than GPT-5.0 despite being significantly more capable.
Enterprise Impact: For businesses, GPT-5.5 means a reduction in "wrapper" code. You no longer need separate OCR, STT, and TTS services. Spud handles the entire pipeline within a single inference call.
How to Access GPT-5.5
GPT-5.5 is currently rolling out to ChatGPT Plus and **Enterprise** users. The API is available via the new `/v2/agents/completions` endpoint, featuring a dedicated "Computer Use" parameter that enables the model's desktop interaction capabilities.
At AI Cortexo, we are already helping our clients integrate Spud into their autonomous workflows. The shift from "AI assistance" to "AI agency" is officially here.