Windsurf: GPT-5.4 Mini Now Available
Windsurf added GPT-5.4 Mini to its model roster on March 17, 2026, available to self-serve users at a promotional starting rate of 1x credits. The model is OpenAI's fastest and most capable compact model to date, running more than 2x faster than GPT-5 Mini while approaching full GPT-5.4 performance on key coding benchmarks β notably scoring 54.4% on SWE-Bench Pro versus 57.7% for the full flagship. Windsurf users gain access to a model optimized for high-throughput coding tasks, agent subworkflows, and multimodal reasoning at significantly lower cost than the full GPT-5.4.
Sources & Mentions
5 external resources covering this update
GPT-5.4 Mini β Hacker News Discussion
Link
GPT-5.4 Mini and Nano
Link
OpenAI releases GPT-5.4 Mini and Nano, its most capable small models yet
Link
GPT-5.4 Mini brings some of the smarts of OpenAI's latest model to ChatGPT Free and Go users
Link
GPT-5.4 Mini is now generally available for GitHub Copilot
Link
GPT-5.4 Mini Arrives in Windsurf
Windsurf expanded its model lineup on March 17, 2026, with the addition of GPT-5.4 Mini β OpenAI's latest compact model and the most capable small model the company has released to date. Available immediately to self-serve users at a promotional rate of 1x credits, GPT-5.4 Mini rounds out the GPT-5.4 family already accessible through Windsurf, giving developers a full cost-performance spectrum for their coding workflows.
Performance That Punches Above Its Weight
GPT-5.4 Mini is positioned as a near-peer to the full GPT-5.4 flagship on the metrics that matter most to developers. On SWE-Bench Pro β the industry benchmark for real-world software engineering tasks β GPT-5.4 Mini scores 54.4%, compared to 57.7% for the full model. That 3.3-point gap is a remarkably small trade-off given the significant difference in speed and cost.
Speed and Multimodal Gains
Beyond coding benchmarks, GPT-5.4 Mini delivers substantial improvements across a broader range of capabilities:
- 2x+ faster than the previous GPT-5 Mini, enabling more responsive interactions and faster agent task completion
- OSWorld-Verified score of 72.1% versus 42.0% for GPT-5 Mini β a dramatic leap in computer-use and multimodal reasoning capability
- Competitive performance on math, science, and instruction-following tasks relative to the full flagship
Pricing and Credit Structure
For Windsurf users, GPT-5.4 Mini is priced at 1x credits under a promotional rate β making it the most cost-accessible OpenAI model currently available on the platform. In the broader OpenAI API ecosystem, the model is priced at $0.75 per million input tokens and $4.50 per million output tokens, substantially undercutting the full flagship while delivering near-equivalent results on the majority of coding tasks.
This positions GPT-5.4 Mini as the go-to choice for:
- High-throughput coding tasks where latency matters
- Cascade subworkflows and background agent pipelines
- Multimodal reasoning tasks involving screenshots or UI interactions
- Cost-conscious teams running large volumes of AI-assisted development
Completing the GPT-5.4 Family in Windsurf
With GPT-5.4 Mini now available, Windsurf offers the full GPT-5.4 tier stack: from 1x-credit Mini tasks at one end to 8x-credit Extra High Reasoning with the full GPT-5.4 at the other. Developers can now select the exact point on the cost-performance curve that fits their use case β whether that is a quick autocomplete, a multi-step refactor, or a deep architectural analysis requiring maximum reasoning depth.