GPT-5.5 ‘Spud’ Launches: OpenAI’s Most Capable Agentic Model Yet

On April 23, 2026, OpenAI released GPT-5.5 — codenamed ‘Spud’ — the first fully retrained base model since GPT-4.5, delivering native omnimodal architecture, autonomous planning capabilities, and state-of-the-art performance across 14 benchmarks.

Key Points:

GPT-5.5 achieves 82.7% on Terminal-Bench 2.0 and 39.6% on FrontierMath Tier 4 — nearly double Claude Opus 4.7’s score on the same test.

The model is natively omnimodal: processing text, images, audio, and video in a single unified architecture, improving visual reasoning alongside coding scores.

GPT-5.5 is Codex-native — it can autonomously navigate browsers, operate software, and execute code to solve unclear problems with minimal guidance.

API pricing has doubled vs. GPT-5.4: $5 per million input tokens, $30 per million output; Pro tier at $30/$180.

Enterprise integrations launched day-one with Snowflake and Databricks, and the model is optimized for NVIDIA GB200 NVL72 rack-scale systems.

Why It Matters:

GPT-5.5 is the clearest proof yet that AI has crossed the ‘passive assistant’ threshold — it does not just answer; it plans, validates, and works through ambiguity end-to-end.

The 46% reduction in errors on Databricks OfficeQA Pro marks the first time AI has demonstrated the reliability required for high-stakes financial and legal workflows at scale.

API price doubling signals the end of the subsidized AI era — organizations should model future costs at 2-5x current rates.

Key Takeaways for AI Enthusiasts:

Give GPT-5.5 high-level goals, not micro-prompts. ‘Analyze these three contracts and flag all liability clauses’ — not 20 individual steps.

If you are building on OpenAI APIs, update your cost models immediately. The pricing shift is permanent and will likely accelerate.

Watch the competitive benchmark battle with Claude Mythos Preview: GPT-5.5 narrowly leads today, but Anthropic has not released Mythos publicly. This race is not over.