OpenAI released GPT-5.5 (codenamed “Spud”) on April 23, 2026. This marks a significant release as it’s the first fully retrained base model since GPT-4.5, moving beyond incremental updates.

Key Features & Capabilities

  • Native Omnimodality: GPT-5.5 processes text, images, audio, and video within a single, unified architecture, enabling more fluid and coherent cross-modal reasoning.
  • Agentic Autonomy: It is designed for complex, multi-step “agentic” tasks, capable of planning, tool coordination, self-verification, and executing long-horizon workflows with minimal human intervention.
  • Advanced Reasoning (“Thinking” Mode): A “Thinking” version utilizes parallel test-time compute for logic-heavy tasks, with a default “medium reasoning effort” setting.
  • Coding & Computer Use: Demonstrates a major improvement in software engineering, resolving 58.6% of real GitHub issues on SWE-bench Pro. It can also operate software in real computing environments for office workflows.

Expanded Context & Efficiency

  • Context Window: Supports up to 1 million tokens for API users (400,000 for Codex).
  • Speed: Token generation is approximately 20% faster than GPT-5.4.
  • Token Efficiency: Uses about 40% fewer tokens for the same tasks, resulting in a net cost increase of approximately 20% despite a higher per-token price.

Performance Benchmarks

GPT-5.5 leads in several evaluations:

  • Terminal-Bench 2.0: 82.7%
  • FrontierMath: 51.7% on levels 1–3 and 35.4% on level 4.
  • MRCR v2 (Long Context): 74.0% accuracy at 1 million tokens.

Availability & Pricing

  • Tiers: Available to ChatGPT Plus, Pro, Business, and Enterprise subscribers, with the “Pro” version for highest tiers.
  • API: Launched on April 24, 2026, priced at 30.00 per 1M output tokens.
  • Ecosystem: Forms the foundation for OpenAI’s new “Super App,” a desktop application integrating ChatGPT, Codex, and a dedicated browser.