Dolphin X1 Trinity Nano is now live on @huggingface Our smallest decensored model yet - 6B MoE with 1B active parameters trained using only online RL Huge thanks to @TargonCompute for providing an 8xB200 node, @PrimeIntellect for hosted RL, and @arcee_ai for the Trinity series

You can download the model today on Hugging Face and run Q8_0 with a 32K context on just 8GB of VRAM It can even run on mobile devices Full weights huggingface.co/dphn/Dolphin-… GGUF huggingface.co/dphn/Dolphin-… FP8 huggingface.co/dphn/Dolphin-…

Node provider rollout has been going well Our pool of inference nodes running Qwen 3.6 35B have generated over 3.2B tokens so far Total inference bandwidth -> 9400 t/s 28x RTX 4090 12x RTX 5090 8x RTX PRO 6000 & many other cards API access coming soon 🐬

Venice Uncensored 1.2 is live The model has been upgraded with vision support, a 4× larger context window & stronger tool-use capabilities Developed in collaboration with @AskVenice to deliver the most uncensored version of Mistral 3.2 24B Trained on Subnet 4 @TargonCompute

This model was trained & decensored with supervised fine tuning followed by a KTO Reinforcement learning run Training was performed on 8xB200 provided by @TargonCompute for over 1 month This resulted in very strong generalized realignment of the model without needing to use any abliteration techniques to directly edit model weights Zero refusals were detected when we benchmarked censorship levels across a collection of 4k commonly refused prompts By using RL instead of abliteration to decensor the model, we avoided the common bugs & performance loss seen after manually editing model weights End result - a model just as smart as Mistral 3.2 24B Instruct with refusals removed across the board