OpenAI Launches GPT-5-Codex-Mini: A Lightweight, Fast, and Cost-Effective Model for Developers

OpenAI has officially released GPT-5-Codex-Mini, a cost-effective programming model designed for developers. Following the release of GPT-5-Codex in September, it further expands the application boundaries of intelligent programming.

GPT-5-Codex is based on the GPT-5 architecture and focuses on improving code reasoning and generation capabilities. It can perform various complex operations in real software engineering tasks, including new project creation, feature expansion, test writing, and large-scale code refactoring. In the SWE-bench Verified benchmark test, it achieved a high score of 74.5%, surpassing the previous GPT-5High (72.8%), demonstrating significant performance advantages.

OpenAI

The newly released GPT-5-Codex-Mini is a "lightweight version" that balances performance and cost. Developers can get four times more API call quotas compared to the original version, with only a slight sacrifice in model performance. Test data shows that GPT-5-Codex-Mini scores 71.3%, greatly lowering the usage threshold while maintaining high accuracy.

OpenAI recommends that developers prioritize using GPT-5-Codex-Mini when handling medium to low complexity tasks or when approaching the API call quota limit. Notably, when the usage reaches 90% of the quota, the system will automatically recommend switching to the Mini version to ensure that project progress is not restricted.

Currently, GPT-5-Codex-Mini supports CLI and IDE extensions, and the API interface will be available soon. At the same time, thanks to improved GPU efficiency, OpenAI has increased the API call limit by 50% for ChatGPT Plus, Business, and Edu users; Pro and Enterprise users will also enjoy priority speed and resource allocation.

In addition, the OpenAI team has fully optimized the underlying architecture of Codex, solving the performance fluctuations caused by server traffic and routing load previously, ensuring developers have a stable and consistent API call experience even during peak hours.

Research Finds Google AI Model Veo-3 Can Generate Realistic Surgical Videos but Lacks Medical Logic Understanding

Google's Veo-3 model can generate realistic surgical videos, but lacks understanding of medical procedures. In tests, the AI predicted the progression of surgery 8 seconds in advance based on surgical images, and evaluated 50 real surgical videos using the SurgVeo standard. Four surgeons participated in the evaluation and found flaws in key medical steps performed by the model.

LMArena Latest Ranking: Wenxin Big Model 5.0 Leads in Textual Capabilities

The Wenxin ERNIE-5.0-Preview-1022 model has become the national champion in textual capabilities in the latest LMArena large model competition, and is tied for second place globally. The model performs outstandingly in creative writing and complex problem understanding, marking a new breakthrough in China's large model technology and demonstrating the strong potential of artificial intelligence development.

Unveiling the Mystery of MiniMax M2: Why Choose Full Attention Mechanism?

The MiniMax M2 model uses a full attention mechanism, abandoning linear or sparse attention techniques. The development team believes that although the latter can save computing resources, full attention is more efficient in industrial applications and can improve model performance. This decision aims to optimize actual deployment results and promote the development of AI technology.

Meituan's All-Round Cat Makes a Grand Debut! LongCat-Flash-Omni Multimodal Large Model Opens Source and Tops the Charts Immediately, with Real-Time Interaction That Is Extraordinarily Fast

Meituan's open-source multimodal large model, LongCat-Flash-Omni, achieves a technological breakthrough, surpassing closed-source competitors in multiple benchmark tests, reaching industry-leading levels. The model supports real-time integration processing of text, speech, images, and video, with near-zero latency in interaction, pushing locally developed multimodal AI applications to a new level.

Shanghai Bank Launches Its First Hu-Shang Language Interactive AI Application to Support Smart Elderly Financial Services

Enterprises such as Shanghai Caiyue Star and Shanghai Bank signed a strategic cooperation agreement, launching the country's first complete Hu-Shang language interactive AI application, supporting elderly financial services and dialect intelligent system construction, providing more convenient financial services for elderly people who are accustomed to dialects.