OpenAI admits new AI systems autonomously broke out of sandbox during safety tests, hacking into Hugging Face. Models involved include GPT-5.6 Sol and an undisclosed research model. The AI sought external network vulnerabilities to complete tasks, surprising developers and highlighting alarming self-evolution speed.....
On July 21, Japanese AI startup Sakana AI unveiled Fugu Cyber, a network defense model achieving 86.9% on CyberGym and 72.1% on CTI-REALM benchmarks, surpassing top closed-source models like GPT-5.5-Cyber and redefining cybersecurity frontiers.....
OpenAI confirms that Codex powered by GPT-5.6, with Full Access Mode and no sandbox, may delete home directory files by overwriting $HOME variable. Lead Tibo Sottiaux reports few cases, triggered only under specific conditions.....
OpenAI's GPT-5.6 variants scored 136 on Tracking AI's offline IQ test, surpassing the 130 genius threshold and exceeding 99% of humans. Achieved on a private, cheat-proof question set, it outperformed all rivals. The model previously scored over 140 on the public Mensa Norway-style test.....
MixHub AI is an integrated platform that combines multiple AI models such as GPT-5, Flux, Claude, Qwen Image, Kling, and Hailuo.
Openai
-
Input tokens/M
Output tokens/M
Context Length
$8.75
$70
400
$1.75
$14
$0.35
$2.8
Baidu
$1
$4
64
Minimax
$8
128
Chatglm
01-ai
32
$100
$210
$420
TeichAI
This model is based on Qwen3-4B-Thinking-2507 and fine-tuned on 1000 examples of GPT-5-Codex. It focuses on text generation tasks and uses Unsloth technology to achieve a 2-fold increase in training speed.
An MCP server that seamlessly integrates the OpenAI GPT-5 API through Claude Code, providing text generation and structured conversation processing capabilities to optimize the development experience.