Good Morning, Model Builders!
Open weights rule the day: Meta drops the 816B Muse Spark Open while NVIDIA ships Nemotron 3.5 Super VL.
Meta's Open-Weight Comeback

TL;DR: Meta released Muse Spark Open, an 816-billion-parameter open-weight model that handles text, images, video and audio with up to one million tokens of context. Meta calls it the most capable US open-weight model yet, and it lands roughly level with the strongest Chinese open models from Qwen, GLM and Kimi on benchmarks. Read more →
Muse Spark Open Runs Locally on Macs

TL;DR: Meta and Apple teamed up so Muse Spark Open runs locally on Apple silicon. Meta says it is the first model in its class that can run on a desktop setup of two M5 Ultra Mac Studios, with MLX optimisation from Apple. For builders, that puts a frontier-class open model on hardware outside the data center. Read more →
NVIDIA Ships Nemotron 3.5 Super VL

TL;DR: NVIDIA released Nemotron 3.5 Super VL, a 120-billion-parameter open-weight multimodal model aimed at always-on agents, claiming the fastest task completion and highest accuracy in its class. The launch also includes the smaller Bolt 74B-A9B model, with an even lighter Bolt E4B variant scheduled for October 23, extending NVIDIA's open-model lineup from data centers to desktops. Read more →
Nemotron Super VL Fits One DGX Spark

TL;DR: Nemotron 3.5 Super VL is a mixture-of-experts model with 120 billion total and 12 billion active parameters, a one-million-token context window and NVFP4 weights that run on a single DGX Spark. Low active parameters and 4-bit weights cut the hardware needed to serve a capable multimodal agent model. Read more →
OpenAI Rolls Out GPT-Live-1.1 Voice

TL;DR: OpenAI launched GPT-Live-1.1 for Voice in ChatGPT and Codex, with new voices, better accents and fewer "checking..." pauses. New voices cover French, German, Hindi, Mexican Spanish, Korean and Japanese. It is rolling out to all users today, and OpenAI says API access is coming soon. Read more →
OpenAI Expands GPT-Rosalind Academic Access

TL;DR: OpenAI is widening access to GPT-Rosalind, its biology research model, through a new Rosalind Academic Plan. Eligible labs at US universities get sponsored access to use the model for understanding biology, exploring ideas and deciding what to test next, bringing the specialised model to more research teams. Read more →
Mistral and CERN Plan Physics Model

TL;DR: Mistral and CERN are joining forces to explore what large-scale AI models can do for particle physics. Their stated ambition is a single Large Physics Model trained on billions of simulated collisions, a domain-specific foundation model rather than a general chatbot. Read more →
Musk Touts SpaceX-NVIDIA Star Mind Computer

TL;DR: Elon Musk said the Star Mind computer being co-engineered by SpaceX and NVIDIA is far better than almost anyone expects and will reach volume production next year. Musk gave no specifications, so capacity, pricing and what it means for Grok training remain unconfirmed. Read more →

Top AI models. Lower costs.
Access GPT, Claude, Gemini, and more through one OpenAI-compatible API—and save up to 70% on AI API costs. Create your DIT account, get your API key, and start building.