OpenAI Ships GPT‑5.5
- OpenAI said on April 23 it released GPT‑5.5, a new flagship model for ChatGPT and Codex that it says is better at coding, computer use, web research, and multi-step work. - OpenAI added GPT‑5.5 and GPT‑5.5 Pro to its API on April 24, pricing GPT‑5.5 at $5 per million input tokens and $30 output, versus GPT‑5.4 at $2.50 and $15. - Databricks said GPT‑5.5 cut errors 46% in its OfficeQA agent test, underscoring how enterprise buyers are benchmarking model gains against governance and cost. (databricks.com)
OpenAI released GPT‑5.5 on April 23, pitching it as a model built to do computer work, not just answer prompts. (openai.com) OpenAI said GPT‑5.5 is rolling out to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex, while GPT‑5.5 Pro is limited to Pro, Business, and Enterprise in ChatGPT. (openai.com) The company said the model is stronger at writing and debugging code, researching online, analyzing data, creating documents and spreadsheets, and operating software across multiple tools until a task is finished. (openai.com) (cnbc.com) OpenAI framed the release as a shift toward “agentic” systems, meaning models that can plan steps, use tools, check their own work, and keep going through ambiguous tasks with less user guidance. (openai.com 1) (openai.com 2) On OpenAI’s published benchmarks, GPT‑5.5 scored 82.7% on Terminal‑Bench 2.0, up from 75.1% for GPT‑5.4, and 78.7% on OSWorld‑Verified, up from 75.0%. (openai.com) OpenAI updated the launch on April 24 to say GPT‑5.5 and GPT‑5.5 Pro were available in the application programming interface, or API, with added safeguards for scaled deployments. (openai.com 1) (openai.com 2) OpenAI’s pricing page lists GPT‑5.5 at $5 per million input tokens and $30 per million output tokens, while GPT‑5.4 is listed at $2.50 input and $15 output. Cached input is $0.50 for GPT‑5.5 and $0.25 for GPT‑5.4. (openai.com) Databricks said it is partnering with OpenAI on GPT‑5.5 and plans to offer OpenAI GPT‑5.5 and Codex through Unity AI Gateway, its governance layer for enterprise model access. (databricks.com) In Databricks’ OfficeQA Pro benchmark, GPT‑5.5 scored 64.66% versus 57.14% for GPT‑5.4 when retrieval was already handled. In its full agent harness, GPT‑5.5 scored 52.63% versus 36.10%, which Databricks said was a 46% reduction in errors. (databricks.com) OpenAI said it tested GPT‑5.5 under its Preparedness Framework, ran targeted red teaming for cybersecurity and biology risks, and gathered feedback from nearly 200 early-access partners before release. (openai.com 1) (openai.com 2) CNBC reported the launch came less than two months after GPT‑5.4, underscoring how quickly OpenAI is updating its flagship models as it competes with Anthropic and Google. (cnbc.com) The immediate test is whether customers pay more for a model that promises fewer hand-holds, better tool use, and stronger results on coding and document-heavy work. (openai.com) (openai.com)
