Inherent, a London-based AI startup, claims its Faraday model surpassed Anthropic's Claude Opus 4.8 and OpenAI's GPT-5.5 at replicating scientific research using a model with just 27 billion parameters.
Netflix's new recommendation system, GenRec, outperformed its existing methods with 1.6 percent better ranking quality in offline tests.
Deepseek's Flash Vision model performs nearly as well as Opus 4.8 on internal agent benchmarks, according to the company.
Anthropic is using its most powerful model, Claude Mythos 5, to enhance cyber security tools for enterprise customers and key industries.
OpenAI announced GPT-Image-2 can now create PNGs with transparent backgrounds via API, improving image quality for complex elements like glass and fibers.
OpenAI's revenue rose 35% this quarter after launching GPT-5.6 Sol, with enterprise revenue up over 50%, according to CNBC.
OpenAI launched a new Apple Messages plugin for ChatGPT, allowing users to manage messages through the chatbot as of August 20, 2026.
Adobe is rolling out three AI audio tools in Firefly, all cleared for commercial use, with Gemini Omni Flash now integrated into the platform.
Amazon Bedrock now supports cross-region inference for OpenAI GPT-5.6 models across more than 25 AWS Regions, offering improved scalability and performance.
LiquidAI today released QAD Q4_0 GGUFs for LFM2.5 models, achieving 97% recovery of BF16 accuracy lost during quantization.
Chinese models like Kimi K3 and GLM-5.3 now match leading US models on most benchmarks, narrowing the gap to a few months.
Anthropic's internal AI model, Model 2, outperforms all publicly available Claude versions, according to its August 2026 Risk Report.