🤖 GPT-4.1 API Release: Performance Improvement, Pricing Adjustment and Model Iteration

🤖 GPT-4.1 API Release: Performance Improvements, Pricing Adjustments, and Model Iterations

OpenAI has released the GPT-4.1 API, with key highlights including performance enhancements, pricing adjustments, and model iterations. The main information is as follows:

* Model Overview: The GPT-4.1 series includes the flagship version, mini version, and nano version, all featuring a 1-million-token context window and excelling in instruction following, tool invocation, and coding.
* Performance Improvements: In the AIME '24 test, GPT-4.1's score increased from 13.1% to 48.1%, a nearly 3.7-fold increase; in the SWE-bench Verified test, the score rose from 33.2% to 54.6%, a 21.4-percentage-point increase. Qodo's tests indicate that GPT-4.1 outperforms Claude Sonnet 3.7 in generating high-quality code reviews. Among 200 real pull requests, GPT-4.1 provided better suggestions in 55% of cases.
* Pricing Adjustments: The input price for GPT-4.1 is $2 per million tokens, and the output price is $8 per million tokens. The mini and nano versions have lower pricing respectively.
* Model Iterations: The GPT-4.5 Preview will be shut down on July 14, 2025. GPT-4.1 will only be available via API, and ChatGPT will gradually integrate the improvements of GPT-4.1.
* Knowledge Cutoff Date: The knowledge cutoff date for GPT-4.1 is June 2024.

(HackerNews)

via Teahouse - Telegram Channel