In Brief: OpenAI released GPT-5.5 on April 23, delivering major benchmark gains and a 60% hallucination reduction over its predecessor — while doubling API prices across the board.
OpenAI shipped GPT-5.5 on Thursday, rolling it out immediately to ChatGPT Plus, Pro, Business, and Enterprise subscribers and making it available to developers through the Codex platform. The release arrives just weeks after GPT-5.4 and marks the fastest major version turnaround in the company's public model history — a cadence that underscores the accelerating tempo of the frontier model race.
The new model posts 88.7% on SWE-bench, the rigorous software engineering evaluation that has become a de-facto industry benchmark, and 92.4% on the MMLU knowledge test. More prominently, OpenAI is highlighting a 60% reduction in hallucination rate compared with GPT-5.4 — a figure targeted squarely at enterprise buyers who have cited reliability as the primary barrier to deeper AI integration in production workflows. Three variants ship simultaneously: GPT-5.5 standard, GPT-5.5 Thinking (extended chain-of-thought reasoning mode), and GPT-5.5 Pro, designed for maximum accuracy on the most demanding tasks.
The pricing structure signals OpenAI's confidence in the capability jump — and its intent to grow revenue from its existing user base. Standard API access rises to $5 per million input tokens and $30 per million output tokens, exactly double the GPT-5.4 rates. GPT-5.5 Pro commands $30 per million input tokens and $180 per million output. Developers routing work through the asynchronous Batch endpoint receive a 50% discount off the standard tier. OpenAI states that despite the performance gains, GPT-5.5 matches GPT-5.4's per-token latency in real-world serving conditions, meaning the cost increase is not accompanied by throughput degradation.
Capability gains are described as most pronounced in agentic coding, autonomous computer use, knowledge-intensive tasks, and early-stage scientific research — the use cases now driving the largest enterprise contracts. The key question for the coming weeks is whether customers absorb the doubled prices or treat them as an inflection point to accelerate testing of rival offerings from Anthropic and Google, both of which have made aggressive moves on pricing and agent tooling in recent months.