AI Model2026
Provides unquantized BF16 weights of Qwen3.6-27B with the base model's MTP head grafted in for high-fidelity, uncensored text (and multimodal) generation. Includes deployment guidance and hardware-tuned variants for A100/H100 and Blackwell-class GPUs.
AI Model2026
Provides multiple GGUF-quantized exports of Carnice V2 (a merged BF16 SFT of Qwen3.6-27B) optimized for llama.cpp and Hermes-style agent traces, with quant tiers targeted at 16–24GB local GPUs and agentic inference.
AI Model2026
GGUF-format, DS4-optimized quantized weights for DeepSeek-V4-Flash, offering q2 (≈80.8 GiB) and q4 (≈153.3 GiB) variants plus an optional small MTP file for speculative decoding. Built for the DS4 inference engine; MIT-licensed.
AI Model2026
Open-source Mixture-of-Experts LLM designed for extremely long-context (up to 1M tokens) text generation and agentic workflows; uses a hybrid attention + MTP design to reduce KV-cache footprint while enabling 42B active parameters and FP8 mixed-precision training.
AI Model2026
Unified omnimodal foundation model for text, image, video and audio understanding and agentic workflows, with support for up to 1M-token context. Combines a sparse MoE LLM backbone, dedicated vision/audio encoders, multi-token prediction, and a hybrid sliding-window + global attention design to reduce KV-cache overhead.
AI Model2026
Provides a GGUF-quantized build of NVIDIA's Nemotron 3 Nano Omni 30B (Reasoning) for local inference — enables multimodal (video/audio/image/text) reasoning, transcription, and document understanding on compatible runtimes such as llama.cpp, Ollama, vLLM, and TensorRT-LLM.
AI Model2026
Multilingual on-device translation model compressed to 1.25-bit via the Sherry quantization, supporting 33 languages and 1,056 directions in a 440MB package for offline mobile translation and demos.
AI Deploy2026
Provides reusable Docker compose files, scripts and benchmarked configs to serve modern LLMs (Qwen, Gemma, etc.) on 1–2 NVIDIA RTX 3090/4090/5090 GPUs. Multi-engine (vLLM, llama.cpp, ik_llama), measured TPS/context tradeoffs, and validated single/dual‑GPU recipes.
AI Model2026
Converts latent representations into high-resolution images by using a conditional pixel-space diffusion decoder that merges decoding and upsampling into a single generative step. Released checkpoints are 4-step distilled (2k and 2kto4k variants) and pair with specific VAE/encoder weights; license restricts use to non-commercial research.
AI Model2026
An instruct-focused LLM (104B total, 7.4B active) optimized for fast, token-efficient inference in agent workflows. Uses hybrid linear attention plus a sparse MoE to raise throughput and cut token use; suited for high-frequency production agents, with some trade-offs in very deep reasoning.
AI Model2026
A trillion-parameter LLM optimized for long-context, low-latency text generation and agentic coding workflows. Combines MLA+Linear Attention and a post-training 'fast thinking' token-suppression strategy to reduce token overhead and improve multi-step execution reliability for production agents.
AI Model2026
Distills DeepSeek‑V4's multi-step structured reasoning into a Qwen3.5‑9B model for fast image-text-to-text reasoning and agentic tool workflows. Trades larger teacher size for inference efficiency and improved procedural reasoning — good for low-latency research, evaluation, and agent integration.