AI Infra2026
Streams routed experts from disk and treats storage, RAM and VRAM as a single inference hierarchy so MoE models (hundreds of billions to trillions of parameters) can run on consumer or heterogeneous hardware; single-file C engines with zero runtime deps and optional Vulkan/CUDA/Metal backends, speed governed by placement and disk I/O.
AI Model2026
A dynamically quantized GGUF build of Ornith-1.5-35B optimized for agentic code-fixing and multi-turn conversations: targets 4-bit/≈22GB deployments, includes a vision projector, a custom importance matrix and a concise chat template.
AI Model2026
Open-weights LLM fine-tuned for phone-based voice agents that prioritizes low latency and reliable tool/function calling. Based on NVIDIA Nemotron 3 Nano (30B total, 3.5B active), supports very long contexts (262,144 tokens) and recommends temperature=0 with thinking disabled for deployment.
AI Model2026
A large open-weights MoE language model for complex coding, long-horizon agentic workflows, and cyber/security evaluations; post-trained from the GLM-5 family with substantial gains over GLM-5.2. Provides FP8/BF16 checkpoints and native support for very long contexts (up to 1M tokens).
AI Model2026
GGUF-quantized, refusal-removed build of Qwen3.8-Flash-Next for llama.cpp that provides multimodal (image+text), reasoning and tool-calling capabilities; released for security research and red-teaming under the Apache-2.0 license.
AI Model2026
A 770B-parameter Mixture-of-Experts instruct model from Tencent that natively supports 1,048,576-token contexts, Gated DSA attention, and speculative MTP decoding; open-sourced under Apache-2.0 with BF16 and FP8 weights for deployable inference.
AI Model2026
Drop-in abliterated (refusal-removed) build of GLM-5.3-Flash that bakes refusal-direction removal into block-FP8 safetensors, yielding an uncensored 320B (18B active) multimodal MoE model with a 1M-token context. Intended for red-teaming, interpretability, and robustness research; MIT license; not for production without added guardrails.
AI Model2026
Provides a cybersecurity-focused CRACK variant of GLM-5.3 FP8 that reduces refusals for offensive-security, red-team, exploit-development and malware-analysis queries while retaining native FP8 speed on Hopper GPUs; MIT-licensed for authorized security work.
AI Model2026
An experimental multimodal model that adds visual understanding to DeepSeek-V4-Flash: accepts text+image inputs and returns text analyses. Improves vision-dependent agent workflows while maintaining comparable text-only performance; released under an MIT license on Hugging Face.
AI Model2026
Sparse MoE causal LLM that uses Mixture-of-Value Attention (MoVA) to store 36B parameters while activating ~4B per token; supports a native 524,288-token context and is released with final checkpoints, training data, and training code under open license.
Studies looping shared transformer layers in Mixture-of-Experts models under matched budgets and proposes SMELT: loop the middle half twice while matching per-token FLOPs, non-embedding parameters, and KV cache. Shows 6.8–18.0% training-FLOPs savings on the compute-optimal frontier, stronger downstream gains on code and long-context tasks.
AI Model2026
NVFP4-quantized checkpoint of Qwen3.8-Flash-Next for GPU-optimized multimodal autoregressive inference — routed MoE experts in W4A4 NVFP4 while attention/ancillary layers remain BF16; ~2.7× smaller than the BF16 source and supports very long contexts.