OpenAI unveils GPT-5.5, claims a "new class of intelligence" at double the API price

The phrase 'new class of intelligence' is doing a lot of work here, and OpenAI hasn't released independent benchmark results to support it. Doubling API pricing while the capability delta remains unverified by third parties is the detail worth sitting with.

OpenAI has been on an aggressive consolidation run. The Codex expansion covered here in mid-April positioned agentic tooling as a competitive wedge against Anthropic, and GPT-5.5 appears to be the model layer that sits above it. That context matters: OpenAI is building a stack, not just releasing a model. The executive departures of Kevin Weil and Bill Peebles, reported around April 17, signaled a deliberate pruning of consumer-facing experiments in favor of enterprise-grade agentic products. GPT-5.5 fits that pattern. What's harder to square is whether the pricing jump reflects genuine capability gains or is partly subsidizing the acquisition spree and infrastructure costs flagged in the tokenmaxxing coverage from the same week.

Watch whether enterprise customers on existing GPT-5 contracts publicly report measurable task-completion improvements on real workflows within the next 60 days. If the agentic gains don't show up in third-party evaluations by then, the 'new class' framing is marketing, not measurement.