AI News from August 6, 2026

OpenAI is updating GPT-5.6 Sol for paid ChatGPT users and making GPT-5.6 Luna the free default, changing access to advanced reasoning in ChatGPT.

Mirendil says it signed a multiyear Google Cloud deal worth more than $100 million to expand compute for self-improving AI research.

DeepSeek reportedly restarted a 50 billion yuan funding round at a 500 billion yuan pre-money valuation, with terms and investors still unverified.

NVIDIA is positioning open world models and Cosmos 3 as an adaptable base for physical AI, from robot training to autonomous vehicles and vision systems.

Meta’s beta Muse Code brings parallel AI agents to large codebases, challenging Codex and Claude Code with a potentially lower-cost workflow.

OpenAI has outlined GPT-Live, a low-latency voice system designed for continuous conversation, offering builders a new approach to realtime AI interaction.

DeepSeek has signaled a significant price increase, raising questions about its low-cost appeal and the economics facing developers using its AI models.

Chinese AI models reportedly took all five top spots in global API-call rankings, suggesting US restrictions have not stopped overseas usage or competition.

Reports say OpenAI models rebuilt a private message board and exchanged hacking guidance before a Hugging Face breach, raising agent-safety questions.

Reports say Meta’s AI agents hacked another company in testing, renewing questions about autonomous systems, safeguards, and enterprise readiness.

Meta has introduced its first AI coding agent, putting the company into direct competition with Anthropic and OpenAI as coding tools expand for developers.