Efficient 4B native-resolution diffusion foundation model for text-to-image generation and instruction-based image editing. Uses a lightweight Mage‑VAE tokenizer and a 4B NR‑MMDiT backbone to produce 512–2048 outputs with low memory and fast inference; ships in Base, RL-aligned and few-step Turbo variants.
Performs instruction-based image editing from reference images using a 4B native-resolution diffusion transformer; the Turbo variant uses 4-step distillation for interactive latency (≈1.02 s per 1024² edit on A100) while supporting semantic, appearance, structure-aware and restoration edits.
4-bit NVFP4 (W4A4) quantized pack of Upstage Solar Open2 250B for vLLM serving on NVIDIA Blackwell GPUs, preserving MoE routing and near-BF16 quality while cutting model size from 500.6 GB to 153.3 GB.
Provides a 2‑bit quantized build of Qwen3.6‑35B‑A3B for local serving via an OpenAI‑compatible HTTP API. Key features: 12.3 GB on disk, eschamoe mixed 2/3‑bit expert quantization with int8 dense layers, runs on a single 16–24 GB NVIDIA GPU and ships with Escha SGLang and ZML runtimes.
Directly maps visual observations and language instructions to continuous robot actions, replacing LLM-centric V→L→A pipelines. Uses separate visual and language encoders with lightweight bidirectional interaction and a compact decoder to cut inference cost and VRAM, achieving ~31 ms latency and <1 GB VRAM on an RTX 4090; suited for real-time robotic manipulation under tight compute budgets.
Provides ComfyUI-ready INT8 MiniMax‑H3 checkpoints (conditioning encoder plus optional generation tail) for a Heretic-edited Qwen3‑VL‑32B source; preserves the vision tower in BF16 and uses row-wise ConvRot INT8 quantization to reduce VRAM needs for ~32GB GPUs. Not a full Transformers generation repository.
Generates complete songs (up to five minutes) from lyrics and a music description, producing 32 kHz stereo WAV with expressive vocals and long-range musical structure. Uses hierarchical LLMs fused with flow-matching/Flow-VAE synthesis for coherent arrangement and timbre; requires CUDA and integrates with Diffusers and SGLang-Omni.
Provides 2-bit quantized weights of Qwen3.8-27B (~10.15 GB) for local deployment, enabling the full 27B parameter model to run on a single 24 GB GPU with long-context support. Delivered as safetensors plus a companion SGLang runtime; measured to match FP8 reference on common benchmarks with small or no quality loss.
Generates low-latency, instruction-driven English and Chinese speech for voice cloning, voice design, and directed performances; supports real-time streaming, reference-free voice creation, and reference-guided cloning. Open-weight PyTorch model released under a research/non-commercial license with GPU recommendations.
Turns past discovery traces into replayable simulators so alternative exploration policies can be evaluated offline ('dreaming'), enabling fast, low-cost meta-level policy improvement for agent-driven discovery across coding, optimization, and GPU-kernel tasks.
Compresses a 27B-class multimodal model into end-to-end ternary weights to run 27B reasoning on-device: 5.9–8.6 GB deployed footprint, 262K-token context, ~98.2% of FP16 benchmark performance; ships MLX and GGUF packs and runs on Apple MLX and CUDA.
A 27B-class language model packaged in GGUF with end-to-end ternary weights for on-device or single-GPU llama.cpp inference; reduces FP16 footprint to ~5.9–7.2 GB while retaining ~98% of baseline performance and supporting up to 262K tokens.