Together AI Launches Together Post-Training Service for Custom Models

🤖 Anthropic may be spending 75% of its computing power building future models Despite reportedly achieving an impressive 85% gross margin in Q3, Anthropic is directing only around 25% of its computing resources toward serving existing customers. The remaining 75% is going toward training future models, post-training improvements and technical experiments. In other words, most of its computing capacity is being used to build what comes next rather than serve today’s demand. Based on the figures cited, Anthropic generates roughly $100 million in annualized revenue per gigawatt of computing capacity across 5 GW. That would imply only 1.25 GW is currently dedicated to inference — the computing power needed to run models for users. The company reportedly plans to add another 10.5 GW in 2027, a massive expansion of its computing infrastructure. @aipost 🏴

Since March, Gemini has an "Import chats" option (Settings & help > Import memory to Gemini). You upload the ZIP from ChatGPT's data export, your old chats show up in Gemini's sidebar with an import icon, and you can carry on in them. Before you do it, this is what Google's help page says: "Your imported and continued chats are saved in your Activity. This data is used to improve our services (including training generative AI models)." So years of ChatGPT chats become Gemini activity under Google's rules, even if you had model training turned off in ChatGPT. It's the whole export or nothing. ChatGPT's export has no way to pick chats, so every conversation goes in, work and personal together. Claude's export at least lets you choose a date range. Uploading a newer export later overwrites th

https://huggingface.co/XHToken/Spark-X2.6-4B-GGUF Introduction A little over a month has passed since Spark-X2.5. In the rapidly evolving world of AI agents, that is enough time to rethink, refine, and move the frontier forward. Building on the Spark-X2.5-4B, we introduce Spark-X2.6-4B , the latest generation of our compact, general-purpose language models for on-device deployment. Spark-X2.6-4B retains exactly the same architecture as Spark-X2.5-4B, achieving its capability gains through improved post-training. It delivers substantial improvements in coding, agent, long-context, multilingual, reasoning, etc. These advances enable more capable and practical on-device AI for developers and users tackling real-world tasks. Technical Highlights : Stronger Coding, Agentic, and Long-Context Cap