GPT-5.5 Now Available in Windsurf Cascade

GPT-5.5 Is Now Available in Windsurf

Windsurf added GPT-5.5 to its Cascade model picker on April 24, 2026, integrating OpenAI's newest frontier model just one day after its public debut. For Windsurf users, this means access to what OpenAI describes as its "smartest and most intuitive" model to date — directly inside the IDE, through the same model selector already familiar from previous GPT-5.x releases.

What Makes GPT-5.5 Different

GPT-5.5 is not an incremental post-training update. It is the first fully retrained base model from OpenAI since GPT-4.5. Every release between GPT-4.5 and GPT-5.5 — including 5.1, 5.2, 5.3, and 5.4 — was built on the same underlying base with additional RLHF and supervised fine-tuning layered on top. GPT-5.5, by contrast, starts from a new foundation, which is why the capability jump is larger than the version number suggests.

OpenAI President Greg Brockman described it as "a faster, sharper thinker for fewer tokens compared to something like 5.4." That efficiency claim is backed by independent benchmarks: GitHub's testing found GPT-5.5 to be 40% cheaper than GPT-5.4 at equivalent output quality, meaning users effectively get more intelligent responses for less token spend.

Performance Benchmarks

On Terminal-Bench 2.0 — the most widely cited benchmark for agentic coding tasks — GPT-5.5 reaches 82.7%, the highest score among all publicly available models. Its long-context reasoning gains are especially pronounced: on the MRCR v2 benchmark at 1M-token context length, GPT-5.5 scores 74.0%, more than doubling GPT-5.4's 36.6%. This is particularly relevant for Windsurf users working in large codebases, where Cascade routinely operates over extensive context windows.

GPT-5.5 also topped the Artificial Analysis Intelligence Index within 24 hours of its release, scoring 60 — outperforming Gemini 3.1 Pro and Claude Opus 4.5 across evaluations.

Capabilities and Use Cases

The model excels at the kinds of multi-step, autonomous tasks that Cascade is designed for: writing and debugging complex code, researching online, analyzing data, and operating across tools until a task is complete. It requires less explicit guidance than its predecessors, meaning Cascade can take longer agent runs with fewer clarifying prompts from the user.

OpenAI positioned GPT-5.5 as a step toward "agentic computing at scale," and the model has already shown strong results in regulated, high-accuracy environments: Bank of New York's CIO cited notably improved hallucination resistance, calling it suitable for institutions that cannot tolerate inaccurate outputs.

Availability and Pricing in Windsurf

GPT-5.5 became available through the Windsurf model picker on April 24, 2026, the same day OpenAI opened API access. The model appears in Cascade's model selector alongside GPT-5.4 and other supported models.

In Windsurf's credit system, GPT-5.5 is tiered by reasoning level:

• Standard variants: 7–32 credits per prompt
• Fast variants: 17.5–80 credits per prompt

For reference, OpenAI's API pricing for GPT-5.5 is $5 / $30 per million input/output tokens — double the rate of GPT-5.4, though the improved token efficiency partially offsets that cost at the task level.

GPT-5.5 is available to Windsurf users on Pro, Max, and Teams plans. As always, the most current pricing and availability can be confirmed directly in the Cascade model picker within the IDE.