LongCat-Flash-Lite-Sparse and Qwen Uncensored Models Released in GGUF
LLMFan46 · reddit · 2026-08-30
Community modeler LLMFan46 released several quantized models, headlined by LongCat-Flash-Lite-Sparse. This 69B-A3B model adds sparse attention and 1M context length support over the original, requiring a custom llama.cpp fork. The release also includes Ultra Uncensored Heretic versions of Qwen3.8-27B and Qwen3.5-122B-A10B, plus Qwen3-Coder-Next and Laguna-S2.1 with vision. Formats include GGUF, Safetensors, GPTQ-Int4, and NVFP4.
More from Infra
- Same 1,000 Cloudflare Durable Objects: $4,050/month cut to $22 with one trick — threepointone · 2026-10-16
- NVIDIA ships FP4-quantized Qwen3.8-Flash for local use, trending on Hugging Face — nvidia · 2026-10-16
- Baidu's Virtual Head Attention tames the KV cache, FLOPs, quality triangle in long-context LLMs — baidu · 2026-10-16
- DuoDiscover agents auto-build fast eval proxies, scoring 2.68x TPU kernel speedup — berkeley_ai · 2026-10-16
- Google Raises 2026 Capital Spending to Up to $205 Billion — emmanuelvivier · 2026-10-16
- Google open-sources AX, a Kubernetes-style orchestrator for billions of AI agent tasks — thisdudelikesAI · 2026-10-16