Descriptions:
Matt Wolfe delivers a benchmark-rich weekly AI news roundup anchored by the launch of GPT 5.5, with enough pricing and performance specifics to serve as a useful reference document. GPT 5.5 is priced at $5 per million input tokens and $30 per million output tokens — exactly double GPT 5.4’s rates — though OpenAI reports it completes equivalent agentic tasks using significantly fewer tokens. On the Artificial Analysis composite intelligence index, GPT 5.5 Extra High becomes the new solo leader, breaking a three-way tie that had existed between Anthropic, Google, and OpenAI heading into the release.
Wolfe walks through the key benchmark comparisons: Terminal Bench 2.0 shows GPT 5.5 at 82.7% versus Claude Opus 4.7’s 69.4% and Anthropic’s unreleased Mythos model at 82%, while SWEBench Pro sees GPT 5.5 underperform at 58.6% compared to Opus 4.7. The episode also covers GPT Images 2.0 in detail, highlighting its thinking capabilities, real-time web search integration for factual grounding, and ability to render dense multilingual text — illustrated with outputs ranging from OS desktop screenshots to magazine-style layouts and anatomical infographics.
Additional context on GPT 5.5’s access tiers (Plus, Pro, Business, and Enterprise in ChatGPT and Codex; API coming soon) and Codex-specific demo use cases rounds out the coverage. For anyone tracking frontier model capabilities and API pricing trends across OpenAI and Anthropic, the episode functions as a well-sourced weekly digest with enough specifics to inform real purchasing and integration decisions.
📺 Source: Matt Wolfe · Published April 24, 2026
🏷️ Format: News Analysis
