Category

Supra-Router-51M

Decides whether a user prompt should be executed locally on an edge small LLM or routed to a larger cloud model, emitting a deterministic pipe-separated decision string. A 51.7M micro-LLM fine-tuned with multi-task sequence generation to predict domain, complexity and code/math flags, optimized for ultra-low latency edge routing.

ThinkingCap: Qwen 3.6 27B

Fine-tuned variant of Qwen3.6-27B that cuts internal reasoning (‘thinking’) token usage by roughly 46% on average while preserving benchmark accuracy and safety behavior. Targets lower latency and inference cost; ships on Hugging Face with GGUF quantizations for local use.

NVIDIA-Nemotron-Labs-3-Puzzle-75B-A9B-NVFP4

Deployment-optimized hybrid MoE LLM (75B total / 9.3B active) produced via Iterative Puzzle compression and Multi-Token Prediction to double server throughput and raise single-GPU concurrency; designed for multilingual reasoning, long-context generation, and high-volume agentic/chat deployments.

DeepSeek-V4-Flash-GGUF

GGUF-format quantized release of DeepSeek‑V4‑Flash for local inference — compatible with llama.cpp and Unsloth runtimes, with guidance for FP4/FP8 mixed precision and Q4/Q8 quantization; tuned for million-token long-context usage.

GLM-5.2 — colibrì int4 container

Provides pre-converted colibrì-format int4 weights so GLM-5.2 (744B MoE) can run by streaming routed experts from disk on a consumer machine with ~25 GB RAM. Includes MTP shard for lossless speculative decoding; requires the colibrì engine and ~400 GB NVMe.

Tess-4-27B

27B multimodal LLM post-trained to prioritize agentic, weight-scaled reasoning over 64K-token contexts. Built on Qwen3.6-27B and released with BF16 weights plus several GGUF quants; aimed at coding, long-document reasoning, tool use and multimodal inspection.

M87 (early-preview)

Enhances KREA-2 Turbo image generations with an aesthetic LoRA trained on a curated 100-image dataset to add stronger composition, richer lighting, softer atmosphere and refined textures; trigger with --preview for art-directed, cinematic outputs in text-to-image pipelines.

Krea 2 Identity Edit

Instruction-driven LoRA fine‑tune for identity‑preserving image edits: give an image plus a plain‑language instruction and it edits pose, outfit, objects or scene while keeping unasked content and subject likeness. Requires the ComfyUI‑Krea2Edit node pack; distributed under the Krea 2 Community License.

LingBot-World v2 — 14B causal-fast

Generates image-to-video world-model outputs using a distilled 14B causal model optimized for chunked, KV-cached inference across long-horizon interactive scenes; offers a real-time 'causal-fast' variant capable of driving near‑real‑time video streams and an agentic harness for action-driven scene synthesis (CC BY‑NC‑SA).

LingBot-Video-MoE (30B-A3B)

Generates videos from text and image+text prompts using a 30B Mixture-of-Experts model tuned for embodied intelligence; includes a refiner and structured prompt rewriter, and supports diffusers/SGLang runtimes with multi-GPU inference.

Qwythos-9B-v2-GGUF

Provides GGUF-quantized builds of the Qwythos-9B-v2 LLM for local runtimes, with multiple quant levels, optional MTP-enabled variants, a 1,048,576-token context window, and an optional BF16 vision projector for multimodal use.

Qwythos-9B-v2

A 9B-parameter Qwen3.5-based multimodal model tuned to preserve chain-of-thought reasoning while eliminating repetition loops; restores native multi-token prediction, supports 1,048,576-token context, and targets research/red-team use.