OpenAI Ships GPT-5.5 With 1M Context and Doubled API Pricing

OpenAI released GPT-5.5 on April 23, rolling the model out to ChatGPT Plus, Pro, Business, and Enterprise subscribers and to Codex users, with the API following within days. It's the third major model release in under two months — following GPT-5.3 and GPT-5.4 — and the first fully retrained base model since GPT-4.5.

OpenAI is positioning GPT-5.5 as its most capable model to date at agentic work: multi-step coding, desktop software operation, and open-ended research. The catch is the price tag, which has doubled.

The pricing shift

GPT-5.5 costs $5 per million input tokens and $30 per million output tokens on the API. That's exactly double the GPT-5.4 pricing, which was $2.50 and $15. The new Pro tier — aimed at the hardest reasoning tasks — runs $30 per million input and $180 per million output.

OpenAI's argument is token efficiency. The company says GPT-5.5 generates fewer tokens to complete the same task, which partially offsets the higher per-token rate. In practice, whether that math works out depends on the workload. For chat and short generations, prices roughly double. For long reasoning traces, token efficiency might bring the net cost closer to flat — if OpenAI's internal numbers hold up in the wild.

Free-tier ChatGPT users did not receive GPT-5.5. That reinforces a trend from the past year: the gap between free and paid ChatGPT is widening, and frontier capability is increasingly paywalled.

What's actually new

OpenAI calls GPT-5.5 "a new class of intelligence for real work" — marketing language, but the engineering changes are real. The model is the first full base retrain since GPT-4.5, meaning it isn't a fine-tune or distillation on top of GPT-5. It's a new foundation.

Context window is now 1 million tokens on both the standard and Pro variants. That matches the top end of what Anthropic and Google offer and, as of today, what DeepSeek's new V4 open-source model ships with. A year ago 200K was considered generous. Now 1M is becoming table stakes for serious enterprise work.

OpenAI emphasizes three capability areas. Coding — writing, reviewing, and debugging code across large codebases. Computer use — operating software, filling forms, driving browsers, and completing multi-step desktop workflows. And research — planning and executing multi-hour research tasks with tool use and citation.

The agent story, finally

GPT-5.5 is the first OpenAI model where the company leans hard into calling it an agent, not a chatbot. The Codex integration is the most visible manifestation: GPT-5.5 inside Codex can take a high-level task, plan it across files, execute terminal commands, and iterate — much closer to Claude Code or Cursor's workflow than the earlier ChatGPT + plugins model.

Why this matters: agentic workloads consume dramatically more tokens than chat. A single "fix this bug across my codebase" task can burn hundreds of thousands of tokens. The doubled pricing makes more sense in that context — OpenAI is effectively rebalancing its economics toward the workloads where the next generation of demand is coming from.

Industry impact

For developers, GPT-5.5 immediately raises the ceiling on what autonomous coding assistants can do. TechCrunch framed the release as a step toward OpenAI's "superapp" vision — where ChatGPT is less a chatbot and more an operating layer for knowledge work. That's aspirational, but GPT-5.5 is the closest OpenAI has come to actually delivering it.

For enterprises, the pricing doubles the cost ceiling on anything built against GPT-5 today. Teams running production workloads on the GPT-5 line will need to decide whether to upgrade for capability, stay on GPT-5.4 for cost, or shop alternatives. Anthropic's Claude Opus 4.7 and Google's Gemini tier now look noticeably more competitive on price.

For the rest of the market, GPT-5.5 resets expectations for what frontier pricing looks like. If the token-efficiency story holds up, others may follow. If not, it creates real pricing pressure room for Anthropic, Google, and DeepSeek to take share.

What analysts and users are saying

Initial reactions from developers on X focused on the coding improvements. Users reported noticeable gains in multi-file refactoring tasks and in the model's ability to stay on-task across very long conversations.

Pricing drew immediate pushback from indie developers and smaller teams, many of whom had migrated to the GPT-5 line precisely because of its February price cut. 9to5Mac reported that the combination of better capability and higher prices has left mid-market SaaS teams evaluating whether to absorb the cost or wait for a cheaper GPT-5.5 Mini variant that OpenAI has hinted at but not announced.

What's next

GPT-5.5 Pro, the higher-reasoning tier, was being rolled out alongside the base model with API access following shortly after. OpenAI has not announced GPT-5.5 Mini or a free-tier version, though both seem likely within the next couple of months based on prior release cadence.

Watch for two things. First, how OpenAI frames the token-efficiency claim as real workloads hit production — the argument lives or dies on empirical numbers. Second, how Anthropic and Google respond. Both have been noticeably quieter on frontier model releases this month. GPT-5.5 raises the pressure on them to ship.

Bottom line

GPT-5.5 is a capability upgrade and a price increase in the same release. For workloads that can use the agentic features — coding, research, multi-step desktop tasks — the math likely still works. For everyone else, this is the moment to re-check whether GPT-5 is still the right default, or whether the competition has caught up enough to matter.