AI News (2026/3/18): OpenAI Launches GPT-5.4 mini and GPT-5.4 nano, Performance Approaches Full Version
Executive Summary:
OpenAI has recently released two lightweight models, GPT-5.4 mini and GPT-5.4 nano. GPT-5.4 mini achieved a score of 54.4% on the SWE-Bench Pro coding benchmark, 3.3 percentage points lower than the full version of GPT-5.4, but it runs twice as fast as the previous generation and supports up to 400,000 tokens of context. GPT-5.4 nano is positioned for ultra-lightweight tasks and costs only 1/12th of the full version.
Details
On Wednesday, March 18, OpenAI announced the release of two new lightweight language models: GPT-5.4 mini and GPT-5.4 nano. These models are designed to provide more efficient and cost-effective solutions for developers with limited resources. GPT-5.4 mini achieved a score of 54.4% on the SWE-Bench Pro coding benchmark, just 3.3 percentage points lower than the full version of GPT-5.4. Additionally, this model runs twice as fast as the previous generation and supports up to 400,000 tokens of context, making it more flexible and efficient in handling complex tasks. GPT-5.4 nano further reduces costs, with a price of only 1/12th of the full version, making it suitable for ultra-lightweight tasks.
Key Points
Performance: GPT-5.4 mini scored 54.4% on the SWE-Bench Pro coding benchmark, 3.3 percentage points lower than the full version of GPT-5.4. This score indicates that despite its smaller size, the model's performance on specific tasks is still close to that of larger models.
Efficiency: Compared to the previous generation, GPT-5.4 mini's runtime speed has improved by 2 times. It also supports up to 400,000 tokens of context, enhancing its flexibility and efficiency in handling complex tasks.
Cost Advantage: The input price for GPT-5.4 mini is only $0.75 per million tokens, and the output price is one-third that of the full version. GPT-5.4 nano's price is even lower, at 1/12th of the full version, significantly reducing the barrier to entry.
AI-ALL In-Depth Commentary
The launch of GPT-5.4 mini and nano by OpenAI not only provides more options for developers with limited resources but also further promotes the普及 (popularization) of AI technology. These lightweight models achieve a good balance in performance, efficiency, and cost, particularly excelling in coding tasks. Although they are slightly inferior to the full version of GPT-5.4 in some aspects, their significant cost advantage and runtime efficiency Make them ideal choices for many small and medium-sized enterprises. Moreover, the ability to support ultra-long context enhances their competitiveness in tasks requiring the processing of large amounts of information. Overall, OpenAI's move is expected to accelerate the application of AI technology in more scenarios and challenge existing productivity tools.
