Cur8 — Tuesday, September 08, 2026
New AI models like llm 0.35 and GPT-6 signal rapid progress in capabilities. OpenAI’s focus on journalism and internal work highlights strategic shifts. Mistral’s €3B raise and vLLM advancements on AMD GPUs show growing industry momentum and infrastructure innovation.
New model releases
llm 0.35 released Sept 7, 2026, adds command-line access to large language models.
GPT-6 Astra generates photorealistic ray-traced scenes and complex simulations from text alone, achieving results previously requiring years of research. It handles niche algorithms, avoids collusion, and improves safety while reducing monitorability.
OpenAI news
More capable, affordable AI expands work capabilities and lowers growth costs, with OpenAI highlighting advancements in model efficiency and reduced computational expenses.
OpenAI expands journalism support with tools, training, and partnerships targeting students, educators, and newsrooms to enhance media production and education.
OpenAI prioritizes rapid development of powerful, aligned AI for defense against AI threats, with Jakub Pachocki emphasizing the need for caution amid urgency.
Big cloud & vendor AI news
vLLM's speculative decoding on AMD MI300X/MI355X GPUs improves output-token throughput by verifying multiple draft tokens per target-model pass, with performance varying by method (e.g., EAGLE-3, DFlash) and proposal length.
Mistral raises €3B at €21B valuation, largest European tech round. Funds will expand open-weight AI infrastructure, supporting 125+ enterprises across 20 countries.
OpenAI, Anthropic, and Google unveiled new models and tools, with updates including advanced language systems, research breakthroughs, and industry partnerships, as tracked in real-time. No specific benchmarks, dates, or pricing were detailed.