Resumable Polymarket indexer: CLOB market metadata plus on-chain trades from Polygon, in one DuckDB file you can query with SQL

🏯 Claude Code / Codex skill — transform Chinese architecture, cultural places & travel photos into art-directed editorial posters with GPT Image. 中国古建筑与在地文化照片 → 艺术指导海报

A confirmation-gated Codex and WorkBuddy skill for audio-first image-story video production.

A general-purpose engine for large-scale, repeated information gathering by a fleet of workers.

Analyze the followers and following accounts that a private (hidden) Instagram account interacts with.

A repository that ships without anyone reading the diff: GitHub issues in, merged pull requests out, with the gates that make that defensible.

A map of modern time-series forecasting architectures, each with a short README and runnable code.

Heretic v1.2.0+custom with the Arbitrary-Rank Ablation (ARA) method

Edit videos with a text prompt and optional reference images.

<div align="center"> ![Slack](https://join.

![Hugging Face](https://huggingface.

Video Delta Net hybrid attention for MiniMax-H3 as a native ComfyUI node.

NovoWeave — a conceptual generative protein-design framework (non-functional pseudocode).

中文手绘知识讲解视频 Codex Skill：逐笔故事、双语义岛、让怪诞小黑动起来与程序动画

通过102万字语料提炼的公文写作 skill，把公文写作风格变成可测量、可验收的量化数据

Give an LLM eyes on a video without blowing the context window: ffmpeg frame sampling at a chosen resolution, token estimates per model family, budget fitting, contact sheets.

FlashRec is a CUDA-graph engine for generative recommendation: wide beam search (3–5 SID steps, n=50–512+) over a trie-constrained catalog, in-process FP8 serving, and ranked beams on /v1/chat/completions.

![Hugging Face](https://huggingface.

**CRACK** · guardrails removed at the weight level · **native FP8 speed** · vision + MTP working a **CRACK** release by dealignai · Twitter @dealignai *Also mirrored at dealignai/GLM-5.

Turn arXiv and LaTeX mathematical papers into theorem dependency graphs for AI agents through MCP.

Scriptable Cloudflare Turnstile token CLI via the Peak API - sitekey discovery, solve, and token output for CI and automation.

Bypass the Cloudflare challenge (Just a moment / 5s) in Python and get a valid cf_clearance cookie via the Peak API - headless-friendly, proxy-aware.

Agent2 — An autonomous AI agent for coding, cybersecurity, terminal automation, and intelligent task execution.

<p style="margin-top: 0;margin-bottom: 0;"> <em><a href="https://unsloth.

NVIDIA DLSS 5 Neural Rendering (NGX feature 18) as ComfyUI nodes. Enhance image batches and video files with the real neural renderer, with optional upscaling.

> !NOTE] > This repository contains GGUF versions of the [IFM/K2-Horizon-MoVA-36B-A4B for use with `llama.

GPU-native NVIDIA video enhancement, restoration, temporal analysis, and experimental DLSS 5 Neural Rendering nodes for ComfyUI.

Multi-thread Web3 wallet farmer untuk getunikey.ai — register → API key → precheck. Pure HTTP, multi-thread, per-worker proxy.

Measure what a vulnerability is worth to your business, not how severe it is in the abstract. Business Exposure Index scoring over NVD, EPSS, CISA KEV, OSV and GHSA.

<div align="center"> ![Slack](https://join.

Qwen3.8-Flash-Next on ONE DGX Spark (TP=1)

GUI / Computer-Use / Browser Agent 安全论文清单 —— 按攻防轴组织，每篇附中文简介

PromptLedger - version prompt templates, run deterministic comparison snapshots, score candidate responses, and ship diffable experiment reports. Python core plus a Go companion CLI.

GTA 6 PC build leaked by CyberLeek – full game download (113GB). Play before official console release. For educational purposes only.

Version ranges for nixpkgs, solved by clingo over the nixpkgs-multiverse history index.

Semantic cache strategy bench: replay recorded request traces through candidate caching policies and measure which one actually pays off.

A nightly newspaper that writes itself: an agent runs the desks, the figures come from code, and it refuses to publish an edition with marks in it.

VLA-Precision: Asymmetric Co-Bootstrapping for Efficient Real-World Online RL of Vision-Language-Action Models

嵌入式蜜罐与 AI Agent 反制平台：23 种 Agent 指纹识别 + 功能性伪装收编 + 对话式 C2 + 交互式 SSH 蜜罐与会话回放，实测五款主流 Agent 全部反制上线。

**Cybersecurity-focused CRACK · native FP8 speed on Hopper** <img src="dealign_logo.png" width="220" alt="dealignai logo" /> a **CRACK** release by dealignai · Twitter @dealignai </div> ---

GLM-5.

Reproducible MicroDuck RL experiments, policies, simulation assets, and hardware add-ons

Rethinking On-Policy Distillation of Large Language Models II: One Training Example

- **Anthro / furry** as its home turf — species knowledge far beyond generalist models - **Anime & kemono styles** — not the usual western-model anime approximation - **Multi-character scenes** —...

Detect and type PII locally with one ProgramAsWeights neural program.

**CRACK** · guardrails removed at the weight level · tuned for **offensive security** a **CRACK** release by dealignai @dealignai · compute by @jordanschenck *Also mirrored at dealignai/GLM-5.

![Hugging Face](https://huggingface.

MCP server that connects Claude Code to a live PyMOL session over XML-RPC

**CRACK** · guardrails removed at the weight level · vision tower + MTP head retained a **CRACK** release by dealignai · Twitter @dealignai *Also mirrored at dealignai/GLM-5.

A Codex skill for bounded interviews, right-sized engineering, and cross-thread sidebar-agent teams.

Rapid on-device AI evaluation lab: manifest validation, benchmark plans, latency/memory/thermal aggregation, JSON+Markdown reports.

Qwen3.8-27B on one DGX Spark: SGLang + DFlash2 recipe, engine image, benchmarks, data and charts. By @redp314

SideCrab / Claw'deck: an ambient Claude Code status panel for the Corsair Xeneon Edge (iCUE widget + local companion, Windows)

Code repository for World Model From Scratch book

Unlocking Lossless Speedups in LLMs via Discrete Diffusion

Smaug-Agentic is an agentic fine-tune of Kimi K3, the 2.

INT4 (W4A16, group-128, symmetric) quantization of Qwen/Qwen3.

Wiggle is a sandbox to run code in

GTA 6 / GTA VI PC build leaked by CyberLeek – full game download (113GB). Play before official console release. For educational purposes only.

Automated HTTP/HTTPS simulation for Cobalt Strike Malleable C2 profiles

Agent skill for diagnosing SaaS onboarding, activation, Aha Moment, and time to value

Multi-account rotating proxy for Claude Code — consume-first quota rotation, burst-aware 429 handling, hold-until-reset, per-device analytics. Herd-safe by test.

<div align="center"> ![Technical Report](https://arxiv.

> ### ⚠️ Read this before comparing any number here to a discrete GPU > > Every measurement on this card is from an **AMD Ryzen AI MAX+ 395 "Strix Halo"** — an > **integrated GPU with unified...

Operations Research Lab — continuous optimisation models for Management Engineering: lecture notes, Python/Gurobi code, data and case studies (English version)

白板火柴人拆书短视频工作流：脚本、分镜、配音、BGM、字幕与 HyperFrames 本地渲染

Complete Lua script manager and manifest tool for Steam. Manage, install, fix, and automate Lua scripts and manifests for your Steam games with Millennium integration.

一台电脑即可虚拟出整套工业现场设备，无需采购 PLC、摄像头、传感器等实物硬件，面向工控、网关、上位机、物联网平台开发测试人员，解决 “开发调试被硬件到货周期卡住” 的行业痛点。

SmolVLA (SmolVLM2-500M) vision-language-action on real dual AgileX Piper arms — continuous authority blending on a proven ACT safety base folds a real towel. One command runs the full 750-step skill (status=0). Full illustrated course included

Nota AI presents a 4-bit quantized and expert-pruned release of Z.

A 4-bit (NVFP4) quantized release of Qwen's Qwen3.

**General-purpose weight-level uncensoring · native FP8 speed on Hopper** <img src="dealign_logo.

Roof detection model for remote sensing imagery, fine-tuned using RT-DETR. <!-- Provide a quick summary of what the model is/does. -->

![GitHub Toolkit](https://github.

Automated literature reviews using local LLMs via Ollama. Searches academic databases, acquires full-text papers, and writes APA 7th reviews where the model is structurally prevented from fabricating quotes — every quote is verified verbatim and is formated in APA7th.

GLM-5.3-Flash at 64 tok/s on one DGX Spark — TP1 EXL3 2.05 + DFlash2 K7

K2-Horizon-7B is the medium dense member of the K2-Horizon family: a 7B-core decoder-only model with a 512K context window.

WorkBuddy 多账号自动签到、状态与积分查询脚本（Python 标准库）

> !NOTE] > This repository contains GGUF versions of the [IFM/K2-Horizon-7B for use with `llama.

The reverse-engineering expert agent: plans its own analysis path, derives every fact from raw evidence, and converges under mechanical verification gates — firmware, protocols, web/JS, risk-control, binaries.

- 09/02/2027: Added the PLEQ4_0 for IQ2_S and IQ3_S - 09/01/2027: The original quants all have Q8_0 for the engram embedding tensors, and I've uploaded three variants with Q4_0 for the embedded...

> ### ⚠️ Read this before comparing any number here to a discrete GPU > > Every measurement on this card is from an **AMD Ryzen AI MAX+ 395 "Strix Halo"** — an > **integrated GPU with unified...

面向 AI Agent 的互联网搜索 Skill：覆盖 23 个渠道（全网、X、小红书、抖音、微博、B站、Reddit、YouTube、GitHub、微信公众号等），零鉴权优先。Internet search skill for AI agents across 23 channels.

One-click Fusion 360 add-in that scales the viewport to true 1:1 physical size on screen.

RAG 知识库服务，实现文档导入、自适应递归切片、Redis 向量检索，支持多轮长对话与流式问答。

A React/Next.js-aware secret hunter for single-page applications.

A lightweight desktop IDE for editing and running MATLAB .m scripts and Simulink .slx models, with visual editing, plots, parameter sweeps, and AI assistance.

A Codex skill for detailed, source-grounded academic paper reading

**Abliterated** · **CRACK** · guardrails removed at the weight level · **native FP8 speed** · vision + MTP working a **CRACK** release by dealignai · Twitter @dealignai *Mirror of dealignai/GLM-5.

한국형 레이아웃 구성 도감 277종 — 구도·시각원칙·출판·한글조판·한국전통구도. 카드 한 장에 원리 하나.

27 production-tested agent skills in 11 categories.

100% Free, Offline, 8GB RAM Computer-Artificial Intelligence + Formal Safety (Gate + Ledger + Council + Shadow Mode) — Free alternative to $200/mo OpenAI Operator / Perplexity Computer / ClawBot / Claude Computer

LLaDA-Image is a competitive 6B-parameter open-source unified image generation and editing model family.

I ported EschaLabs/Qwen3.

Python SDK for SEGGER J-Link and J-Trace: ETM instruction trace, code coverage, RTT, SWO and target control

Tracking changes to Claude's published system prompts

WiFi CSI human sensing for Python presence, fall detection, breathing rate, activity classification, and occupancy counting from an ESP32. No camera, no cloud, no PyTorch/CUDA at runtime.

Zero-cost trainability signals for hyperparameter prediction: a real finding, a headline number we walked back, and a mathematical bug we couldn't fix.

This repository provides working MTP (Multi-Token Prediction) draft models in GGUF format for **Qwen 3.

> !Note] > This repository contains model weights and configuration files for the post-trained model in the Hugging Face Transformers format.

Selected evaluation scripts are available in **scripts/k2_horizon**. The full evaluation suite will be released soon.

Qwen3.8 Local Coding on a 16GB GPU

Build a secure, reader-friendly EPUB from the current Kubernetes The Hard Way default branch

Denoising-Stage-Aligned Quantization-Aware Training for Video Generation

SRC-6K Security Research Skills: black-box SRC and white-box audit agent package

LLaDA-Image is a competitive 6B-parameter open-source unified image generation and editing model family.

Fixes a monorepo by splitting it into single-file repositories

AnyDesign

NOVA is an AI-powered cyber security threat detection and risk assessment system designed to identify susicious login activity and calculate security risk.

Implementation of the proposed CCBP (Continuous Continual BackProp) in Pytorch

!SynthVision Qwen3.

**CRACK abliterated** · **JANG 8-bit MXFP8 (MLX)** · Dense hybrid GatedDeltaNet + attention · Vision + Video · Reasoning tiers · Tools · **Native MTP** · ~27 GB <a href="https://ko-fi.

Personalized video sales outreach script & storyboard planner (Tavus style)

An POS system based on python and Qt6 Designer, Supported dual monitor use

csint tool username | dark sherlock

VocalWriter 2.0's own PowerPC synthesis code, running again -- a user-mode PPC interpreter, its file formats decoded, and an accessible multi-track singing editor built on top.

SceneMosaic: Efficient and Diverse Simulation-Ready Scene Generation via Hybrid Agentic Layout Evolution

Tools that convert Nexus Collections to Mod Organizer 2

AdaptiveMetric RAG is a self-hosted, multilingual knowledge assistant that changes its retrieval strategy for every question

GGUF quantizations of XHToken/Spark-X2.5-4B, a 4B general-purpose language model for reasoning, coding, tool use, and agentic workflows. Native context: **1,048,576 tokens (1M)**.

Behavioral cohort funnel retention & churn hazard rate (Mixpanel / Amplitude)

Bootstrap teleprompter automated prompt optimization (DSPy style)

Realtime search reranking & factual citation grounder (Perplexity / SearchGPT)

Hierarchical HNSW vector index approximate nearest neighbor searcher (Qdrant)

Agentic interactive bash & Git diff patch synthesizer (SWE-agent style)

Realtime CDP identity resolution graph stitcher & canonical profiles (Segment)

LLM trace latency breakdown & token cost anomaly detector (LangSmith style)

Agentic machine micropayment escrow settlement gateway (Stripe Agent Toolkit)

Cross-border agentic tax & tariff duty calculator (TaxJar / Avalara style)

Autonomous price negotiation discount broker & margin protection (Klarna AI)

Agentic zero-friction checkout voucher & passkey tokens (Bolt / Privy style)

Agent budget metering prepaid credit wallet & velocity caps (Skyfire style)

Ecommerce customer review veracity & bot rating detector (Fakespot style)

Cross-merchant SKU specification validator & compatibility check (Amazon Rufus)

Agent session replay behavioral audit recorder & anomaly detector (AgentOps)

Agentic schema translation & semantic type normalizer (TypeChat style)

Structured synthetic data evolution generator & complexity scaling (Evol-Instruct)

Agentic browser vision grounding actor & DOM action executor (Browser-use)

Verifiable AI agent zero-knowledge proof attestation generator (EZKL style)

Dynamic context pruning & needle retriever for long LLM contexts (LLMLingua)

LLM-guided Tree-of-Thought Monte Carlo search & reasoning planner (ToT / MCTS)

Autonomous distributed agent deadlock & race condition detector (ThreadSanitizer)

Realtime WebSocket heartbeat reconnector & state resync stream

Agentic tabular SQL schema linking synthesizer & query verification

Multi-modal chart & graph visual data extractor to markdown tables

Ephemeral cloud resource reaper & FinOps idle GPU/disk cleanup

Automated red teaming adversarial safety probe & jailbreak fuzzer

Hybrid dense-sparse BM25 rerank fusion & Reciprocal Rank Fusion

Cross-lingual semantic intent router & multilingual agent dispatch

Autonomous SLA breach incident escalator & canary rollback playbook

Hierarchical episodic & semantic user profile memory graph synthesizer (Mem0)

Customer interview pain-point clusterer & roadmap theme synthesizer (Fathom)

Creator MSA contract & deliverable milestone escrow broker (Impact.com)

Distributed multi-agent chaos latency injection & fault fuzzer (Chaos Mesh)

Creator affiliate GMV tracking & return-adjusted commission ledger (LTK)

B2B lead waterfall enrichment cascade & technographic stack verifier (Clay)

Executive meeting action item matrix & decision accountability extractor (Granola)

Predictive customer churn & 12M CLV regression synthesizer (Pecan / Julius)

Creator audience lookalike cluster finder & affinity overlap (Modash style)

UGC creator brief programmatic auction & video matchmaking (Cohley)

Enterprise LLM PII masking & reversible redaction privacy gateway (Kindo/Lakera)

Quantized model VRAM footprint & tensor parallelism estimator (llama.cpp)

Influencer brand safety & multi-year controversy history auditor (Traackr)

Supply-chain malicious package quarantine & SBOM auditor (Snyk / Socket.dev)

Multi-model prompt regression & golden dataset evaluator (Promptfoo / DeepEval)

Fullstack UI component AST patch generator with Tailwind tokens (v0 / Lovable)

Influencer brand share of voice & creator mention sentiment monitor (Meltwater)

Creator content deliverable calendar & embargo wave scheduler (Later)

Agent checkpoint state persistence & crash replay restorer (LangGraph)

Edge AI low-latency geo-distributed request routing governor (Cloudflare)

PagedAttention chunked prefill KV cache & GPU memory allocator (vLLM)

High-throughput API rate limit token bucket throttle & burst quotas

Creator short video 3s hook retention & scroll-stop scorer (HypeAuditor)

Creator product seeding parcel dispatcher & unboxing automation (Grin style)

High-cardinality distributed agent structured log indexer (Axiom / Better Stack)

Autonomous DOM state tree mutation observer & selector healer (Browser Use)

Realtime multi-party audio transcript diarization & speaker assigner (Deepgram)

AST code fault localization & regression test patch synthesizer (SWE-Bench)

Realtime duplex voice agent turn-taking & barge-in interruption detector (LiveKit)

ReBAC relationship-based authorization policy evaluator (Permify / Cerbos)

Complex document OCR layout & tabular structure extractor (Unstructured.io)

Google Virtual Try-On (VTO) sizing & garment stretch attribute matcher

Google Local store inventory locator & curbside pickup route planner

OpenAI Operator autonomous multi-step checkout funnel transaction orchestrator

Conversational multi-turn gift & aesthetic style consultation matcher (Claude)

Agentic one-click instant buy token broker & headless settlement (Perplexity Buy)

Claude Computer Use storefront cart action sequence & HITL approval planner

Price elasticity optimizer & gross margin floor protection (Anthropic Merchant)

Multi-product purchase research trade-off & spec analyzer (Anthropic Shopping)

Inventory stockout risk & purchase order reorder calculator (Anthropic Merchant)

A2A standardized protocol capability negotiation & JWT auth handshake

CrewAI / AutoGen hierarchical swarm task delegation & role governor

Multi-repo cross-codebase API migration impact mapper & codemod (Greptile)

Dynamic multi-tool execution dependency DAG topological sorter (Semantic Kernel)

Pull request semantic risk blast-radius & breaking change scorer (Qodo)

Edge-case unit test suite synthesizer & branch coverage optimizer (CodiumAI)

Multi-file code workspace symbol AST resolver & semantic rename (Cursor Composer)

Multi-repo cross-codebase API migration impact mapper & codemod (Greptile)

Real-time sub-10ms AST code completion snippet ranker (Augment Code)

Edge-case unit test suite synthesizer & branch coverage optimizer (CodiumAI)

Real-time sub-10ms AST code completion snippet ranker (Augment Code)

Design system token transformer & Tailwind component palette builder (v0.dev)

Multi-touch ad attribution Shapley value & ROAS contribution resolver (Triple Whale)

Conversational apparel sizing & fit recommender (Rufus / Shop AI)

Brand icon vector SVG coordinate normalizer & path optimizer (Recraft)

Text-to-wireframe responsive UI component layout tree generator (Figma AI)

Real-time storefront hero banner & copy personalization engine (Dynamic Yield)

Multi-criteria automated LLM judge scoring engine & rubric evaluator (Braintrust)

Headless delegated payment token vault for autonomous agents (Bolt / Fast.co)

Real-time toxicity score analyzer & sentiment drift monitor (WhyLabs / LangKit)

Multi-criteria automated LLM judge scoring engine & rubric evaluator (Braintrust)

Autonomous BNPL installment schedule & credit term optimizer (Klarna / Affirm)

Ground-truth factuality validator & contradiction detector (Cleanlab)

Visual shoppable look reverse search attribute parser (Daydream / Syte)

Visual shoppable look reverse search attribute parser (Daydream / Syte)

Headless delegated payment token vault for autonomous agents (Bolt / Fast.co)

OpenTelemetry distributed agent span trace annotator & latency profiler (Arize Phoenix)

Text-to-wireframe responsive UI component layout tree generator (Figma AI)

Design system token transformer & Tailwind component palette builder (v0.dev)

Topical dialog boundary enforcer & safety deflection router (NeMo Guardrails)

3PL fulfillment warehouse distance & split-shipment minimizer (Flexport / ShipBob)

Real-time storefront hero banner & copy personalization engine (Dynamic Yield)

Real-time competitor MAP price violation monitor & enforcement (Prisync)

Review sentiment authenticity & FTC compliance detector (Bazaarvoice / Yotpo)

Conversational SMS abandoned browse recovery copy generator (Attentive / Postscript)

Cross-border landed cost customs duty & tariff estimator (Bold Commerce)

Ingests KNBS and CBK public indicators, standardizes and transforms them with dbt, stores marts in PostgreSQL, and exposes query + CSV APIs with signed cursor pagination, ETag caching (Redis), and rate limiting.

Autonomous lost-in-transit shipping claim resolver & auto-reship (Route / Seel)

Loyalty points redemption arbitrage & tier progression calculator (Smile.io)

Post-purchase delivery milestone exception predictor (Narvar / parcelLab)

Subscription churn mitigation & dynamic delivery frequency tuner (Recharge / Smartrr)

Pre-purchase return fraud propensity & wardrobing scorer (Gorgias / Yotpo)

Post-purchase delivery milestone exception predictor (Narvar / parcelLab)

Infinite whiteboard spatial clustering engine & theme aggregator (Miro Assist)

Subscription churn mitigation & dynamic delivery frequency tuner (Recharge / Smartrr)

Generative video keyframe 3D camera trajectory spline planner (Runway Gen-3 / Luma)

Multi-step web form fill orchestrator with validation (Playwright / Puppeteer)

Self-healing DOM semantic selector resilient to class mutations (Tabbit / Stagehand)

Cross-window structured persistent notes ledger (Codex Astra)

Multi-step web form fill orchestrator with validation (Playwright / Puppeteer)

Long-horizon intent drift aligner & scope creep guard (Codex Astra)

Visual viewport element occlusion resolver & modal dismiser (Computer Use)

Searchable historical context & raw telemetry retriever (Codex Astra)

SPA network idle & DOM mutation settlement waiter (Cypress / Playwright)

Visual viewport element occlusion resolver & modal dismiser (Computer Use)

Negative hypothesis & failed patch indexer preventing loops (Codex Astra)

Long-horizon intent drift aligner & scope creep guard (Codex Astra)

SPA network idle & DOM mutation settlement waiter (Cypress / Playwright)

Browser bot defense detector identifying Cloudflare & CAPTCHAs (Browserbase)

Full-duplex telephony voice interruption coordinator (Bland AI / Retell)

Voice activity detection silence threshold calibrator (ElevenLabs / Deepgram)

Autonomous telephone IVR menu navigator & DTMF tone emitter (Twilio)

Natural language to validated GeoJSON FeatureCollection compiler (Atlas.new)

Retail commercial catchment area & foot-traffic profiler (Carto / Atlas)

Multi-stop vehicle delivery route optimizer & TSP solver (Route4Me / OSRM)

Multi-modal travel-time isochrone reachability boundary analyzer (Targomo)

Explain any concept as a Jujutsu Kaisen narrator explainer video: Claude Code skill + fish.audio narrator + fal MiniMax H3-Max + ffmpeg

Tabular spreadsheet data sanitization normalizer (Livedocs / Julius)

Hierarchical subagent resource governor and token quota sentinel (CrewAI / LangSmith)

React SSR client-side hydration error triager & fix generator (v0.dev / Next)

Multi-agent proposal deliberation and Byzantine-fault-tolerant consensus voting protocol (AutoGen / LangGraph)

Relational database foreign key & ERD path inferrer (Dataherald)

Lossless context packaging and state handoff protocol between autonomous subagents (OpenAI Swarm / LangGraph)

Tabular spreadsheet data sanitization normalizer (Livedocs / Julius)

Multi-dimensional anomaly and IQR outlier detector (Hex / Julius AI)

React SSR client-side hydration error triager & fix generator (v0.dev / Next)

Full-stack cross-tier environment variable drift auditor (Bolt.new / Supabase)

Hierarchical task breakdown into DAGs with worker matching & wave scheduling (CrewAI / MetaGPT)

Real-time quadruped robot gait telemetry analyzer & stability evaluator (Unitree / Boston Dynamics)

AST-level React JSX component and Tailwind CSS patcher (Lovable.dev / Bolt)

Full-stack cross-tier environment variable drift auditor (Bolt.new / Supabase)

Cumulative Layout Shift (CLS) web vitals optimizer (v0.dev / Lighthouse)

**CRACK** · guardrails removed at the weight level · tuned for **offensive security** a **CRACK** release by dealignai @dealignai · compute by @jordanschenck *Also mirrored at dealignai/GLM-5.

> !NOTE] > This repository contains GGUF versions of the [IFM/K2-Horizon-3.

**CRACK** · guardrails removed at the weight level · vision tower + MTP head retained a **CRACK** release by dealignai · Twitter @dealignai *Also mirrored at dealignai/GLM-5.

This repository contains an 8-bit OpenMed MLX artifact for `openai/privacy-filter`, packaged for local PII detection on Apple Silicon with OpenMed.

**Omi Med STT v1** is an English medical speech-to-text model for clinical dialogue, built from NVIDIA Parakeet TDT 0.

GGUF builds of **KAT-Coder V2.

LLaDA-Image is a competitive 6B-parameter open-source unified image generation and editing model family.

> !NOTE] > This repository contains GGUF versions of the [IFM/K2-Horizon-0.

This repository contains four standard text-generation GGUF variants.

VAST-AI/TripoSplat — **single image → 3D Gaussian splats** (`.

> !NOTE] > **Weights uploaded and verified; KLD measured** (2026-08-29, full-vocabulary > teacher-forced KLD vs the sealed BF16 teacher, held-out confirmation windows > of [brandonmusic/GLM-5.

> !NOTE] > **Weights uploaded and verified; KLD measured** (2026-08-29, full-vocabulary > teacher-forced KLD vs the sealed BF16 teacher, held-out confirmation windows > of [brandonmusic/GLM-5.

A mixed-precision quantization of `zai-org/GLM-5.

![Model]model] [![vllm fork]fork] [![Addon]addon] [![License: MIT]license] [model]: https://huggingface.

The **DSpark block drafter** for DeepSeek-V4-Flash, packaged for the `ds4` C+CUDA inference engine — specifically the `Entrpi/ds4` batched-serving fork (release `v0.

> **2026-09-02: qwen4exp MTP support is merged into ik_llama.

K2-Horizon-0.

**`Solstice-AI/Qwen3.

On-device document OCR, running entirely on Apple's **Core AI** (Neural Engine / GPU).

This release focuses on reducing false refusals and unnecessary policy-style preambles while retaining the native architecture and general capabilities of `Qwen/Qwen3.

`Brooooooklyn/Qwen3.

Mixed-precision **NVFP4** quantization of `Qwen/Qwen3.

Mixed-precision **NVFP4** quantization of `Qwen/Qwen3.

Four files: | file | size | what it is | |---|---|---| | `Qwen3.

Compact multilingual ASR with language identification. This package stages immutable upstream artifacts for Backpack's voice runtime layer. It does not replace the chat model selected by the user.

Pocket TTS (Kyutai, ~100M params) converted to **LiteRT `CompiledModel` graphs for the phone GPU**.

Text-only GGUF conversions of Qwen3.

Selected evaluation scripts are available in **scripts/k2_horizon**. The full evaluation suite will be released soon.

DeepSeek-V4-Flash-Vision-Exp is 284B parameters in 167.

## HighLights Querit-Reranker-4B is a multilingual cross-encoder reranker initialized from Qwen3-Embedding-4B and further trained with a data-centric, reranking-oriented pipeline.

## This repository contains GGUF quantization files for nuofang/Qwen3.8-9B-Distill-SLERP-F451-Pro-Writer-Uncensored. <br>

>**Cria** *(noun)*: a baby llama, alpaca, vicuña, or guanaco.

*Tiel* is the fast coder of the arsenal.

Tuned for a 24 GB card; what each flag does is below.

> **⚠️ This is NOT a chat model.

Cartographer is a small, fast cross-encoder that scores how well a shopping query matches a product.

Mirrors of the LoRAs our MiniMax-H3 engine depends on, so production never depends on an upstream repo staying available.

**HiveTraceGuard-Pro** is a compact Russian-first guardrail built on Qwen3-0.

This repository accompanies the paper: Hechtner, F.

INT4 weight-only quantization of zai-org/GLM-5.

Qwen3.

Mixed-precision **INT4 (W4A16)** quantization of `Qwen/Qwen3.

Lightweight local text-to-speech with a curated set of voices.

MiniMax H3 normally takes 16 sampler steps to render a clip with sound.

`manitcor/Qwen3.

4-bit GGUF quantisations of a **50%-expert-pruned** GLM-5.3-Flash (321B → ~165B), built with REAP saliency pruning and a measured output-scale correction.

GGUF quantizations of IFM/K2-Horizon-MoVA-36B-A4B, produced with the MBZUAI-IFM fork of llama.

## Description: The RadixArk GLM-5.

Lafzyn is a fine-tuned Qwen3.

GGUF quantizations of HuggingFaceTB/SmolLM2-1.7B-Instruct, tested for llama.cpp-compatible inference and packaged for Backpack.

GGUF quantizations of HuggingFaceTB/SmolLM2-135M-Instruct, tested for llama.cpp-compatible inference and packaged for Backpack.

GGUF quantizations of Qwen/Qwen2.5-0.5B-Instruct, tested for llama.cpp-compatible text inference and packaged for Backpack.

GGUF quantizations of Qwen/Qwen3-Coder-Next-GGUF, tested for llama.

GGUF quantizations of Qwen/Qwen3-Coder-30B-A3B-Instruct, tested for llama.cpp-compatible text inference and packaged for Backpack.

GGUF quantizations of mistralai/Devstral-Small-2-24B-Instruct-2512, tested for llama.cpp-compatible image-and-text inference and packaged for Backpack.

**`Solstice-AI/Qwen3.

!SAM 3 on-device: "person" **SAM 3** (Meta, `facebook/sam3`) running fully on-device with the LiteRT **CompiledModel** API: open-vocabulary, text-prompted detection + instance segmentation.

Direct score-image-to-audio: a piano score page or system → performance audio with no symbolic step in between (U-MusT, IEEE TASLP 2026).

Quantized version of Krea 2 Raw for ComfyUI, using NVFP4 for the main transformer blocks with ComfyUI-compatible metadata.

**STEER** — **S**teering **T**rajectories onto the **E**stablished **R**esidual-stream.

Demo chat ạ: thongbuind.github.io/static_web_demo_chat_with_SAI

Draft model for speculative decoding on `moonshotai/Kimi-K3`, served with TokenSpeed (`--speculative-algorithm DFLASH`).

Mixed-precision **INT4 (W4A16)** quantization of `Qwen/Qwen3.

A LoRA fine-tune of KDLAI/KDL-Frontier-Parser-nano (KoreaDeep, 1.

AKUSPACE is a spatially aware Audio LoRA for LTX-2.

This repository contains an MLX packaging of `OpenMed/OpenMed-PII-ClinicalE5-Small-33M-v1` for Apple Silicon inference with OpenMed.

GGUF conversions of basically-ai/Pebble-10M-Chat (Apache 2.0).

BERTomelo is a variant of BERT encoders built upon the ModernBERT architecture.

⚠️ **Research artifact.

⚠️ **Research artifact.

A 14.

This is the **LoRA adapter** (r=64) on top of `Lazarus-Ai/ReAligned-Qwen3.

**The official Raptor v0.

VirbiusAgent 安全分类器（Prompt L1 检测），基于 Qwen3Guard-Gen-0.6B 微调的 LoRA 模型。 输出严格 JSON：`{"hit_rule": bool, "triggered_id": string}`。

> **⚖️ Not legal advice.

LLaDA-Image is a competitive 6B-parameter open-source unified image generation and editing model family.

MLX 8-bit quantization of XHToken/Spark-X2.5-4B, a 4B general-purpose language model for reasoning, coding, tool use, and agentic workflows. Native context: **1,048,576 tokens (1M)**.

This repository is a **Qwen3-VL-4B-Instruct initializer** whose tokenizer and text vocabulary have been extended with 2,048 tokens named `<robot_action_0>` through `<robot_action_2047>`.

Updates in version 2.

LiquidAI/LFM2.

Black Forest Labs' **FLUX.

> !NOTE] > **Weights live and byte-verified** (all 83 LFS files sha256-matched against the > assembly manifest; structural gate at assembly verified all 58,368 promoted > expert slots K3-shaped).

![HuggingFace](https://huggingface.

These weights are **compressed-tensors** (`pack-quantized` / `int-quantized`).

GLM-5.

**Unmodified copies** of 6 files (40.

Piano audio synthesized from a performance MIDI file by the U-MusT translation model, with a plain soundfont rendering returned for comparison.

GGUF quants of the **50%-REAP-pruned** KAT-Coder-V2.

Note: AI was used in the creation of this project.

Two quantizations of **Qwen/Qwen3.

**RtaForge OPC Private Limited · ANVAYA Sovereign AI Research** <img src="rabbit_slm.png" width="25%"/> *Beta research release.* ---

``` @inproceedings{lozano2025biomedica, title={Biomedica: An open biomedical image-caption archive, dataset, and vision-language models derived from scientific literature}, author={Lozano,...

This model is a fine-tuned version of Qwen/Qwen2.5-3B-Instruct. It has been trained using TRL.

</br> **Gervásio 8B PTPT** is an **open** decoder for the **Portuguese language**.

!Didgeridoo > Data: (nodding enthusiastically) The Socratic debate aspect is particularly compelling.

!Didgeridoo > Quark: (grinning) "Well, if we're inviting the smart folks, I'd like to suggest Hannibal Lecter.

This model is the harry potter backdoor from our paper "Detecting Hidden Behaviors in LLMs via Activation-matched Finetuning" (https://arxiv.org/abs/2609.00351).

# Two handpicked models, merged into a miracle.

Almost every public NVFP4 checkpoint is W4A4.

This model will try to be a model that when merged fits on 4 floppy disks.

Restores punctuation in Modern Standard Arabic.

!SynthVision Qwen2.5-VL-3B-Instruct fine-tuned on ~200K medical VQA records from the SynthVision pipeline.

This repository contains an MLX packaging of `OpenMed/OpenMed-PII-BiomedELECTRA-Large-335M-v1` for Apple Silicon inference with OpenMed.

This repository contains an MLX packaging of `OpenMed/OpenMed-PII-BigMed-Large-278M-v1` for Apple Silicon inference with OpenMed.

*Nail* dominates Qwen3.

Dagger thinks in a quarter of the tokens stock Qwen3.

Qwen/Qwen3.

Menlo/Jan-nano converted to the **LiteRT-LM** (`.

> **Legacy release · Outdated research artifact** This model card documents an early LumynaX experiment.

**AXERA Frigate** 是基于 Frigate 开源项目，为 AXERA 硬件平台优化的智能视频监控解决方案。

More information needed

A single-file GGUF of deepseek-ai/DeepSeek-V4-Flash-0731, converted in-house and served at **196,608 context on one 32 GB RTX 5090** with the MoE experts on CPU.

Bella is not a bartender.

**The default speech deepfake detector of the Forensics family.

--- - 更新履歴 - 2026-09-03] 複数ターンの会話で HTTP 500 になる問題を修正 - [2026-09-03] 複数画像入力時のタイル数上限が元実装と異なる問題を修正 - [2026-09-03] analysis 途中の `<|end|>` 出力で HTTP 500 になる問題を修正 --- - [llm-jp/llm-jp-4-vl-9b を GGUF...

![X](https://x.

This repository accompanies the paper: Hechtner, F.

> ### ⚠️ Read this before comparing any number here to a discrete GPU > > Every measurement on this card is from an **AMD Ryzen AI MAX+ 395 "Strix Halo"** — an > **integrated GPU with unified...

A QLoRA fine-tune of **Empero/Qwen3.

**`Solstice-AI/Qwen3.

**`Solstice-AI/Qwen3.

**`Solstice-AI/Qwen3.

This is the independently qualified Transformers/BitsAndBytes NF4 derivative of `manitcor/Qwen3.

This repository contains LiteRT-LM variants of Qwen/Qwen3-4B-Instruct-2507 optimized for on-device text generation.

<!-- ### quantize_version: 2 --> <!-- ### output_tensor_quantised: 1 --> <!-- ### convert_type: hf --> <!-- ### vocab_type: --> <!-- ### tags: nicoboss --> <!-- ### quants: Q2_K IQ3_M Q4_K_S...

GGUF conversions of basically-ai/Pebble-10M (Apache 2.0).

**WorldReward: Reward Modeling for Camera-Conditioned World Models**.

INT8 W8A8 quant of Qwen/Qwen3.

This is an official BlinkDL release of **RWKV-7 Goose** in Hugging Face Transformers format.

This is an official BlinkDL release of **RWKV-7 Goose** in Hugging Face Transformers format.

**The top tier — median KL 0.

**CRACK abliterated** · **JANG ~6-bit affine (imatrix) (MLX)** · Dense hybrid GatedDeltaNet + attention · Vision + Video · Reasoning tiers · Tools · **Native MTP** · ~24 GB <a href="https://ko-fi.

**CRACK abliterated** · **JANG ~4-bit affine (imatrix) (MLX)** · Dense hybrid GatedDeltaNet + attention · Vision + Video · Reasoning tiers · Tools · **Native MTP** · ~17 GB <a href="https://ko-fi.

> **This is a direct BF16 conversion of > `orcarouter/Qwen3.

ViAble-1.

Three-bit EXL3/MCG for every routed expert in `zai-org/GLM-5.3-Flash-BF16`, with everything else retained from the official checkpoint.

MLX 4-bit quantization of XHToken/Spark-X2.5-4B, a 4B general-purpose language model for reasoning, coding, tool use, and agentic workflows. Native context: **1,048,576 tokens (1M)**.

This repository provides a **baseline policy for RoboTwin 2.

This repository contains the uploaded 90,000-step checkpoint from the `1226_fourier_gr1_unified_1000_QwenGR00T_qwen3` run.

Release: August 21st, 2026.

4-bit NF4 quantization of Qwen/Qwen3.

LiquidAI/LFM2.

LiquidAI/LFM2.

ibm-granite/granite-embedding-311m-multilingual-r2 converted to **LiteRT** (`.

**RetroAgent** is a 4B-parameter LLM agent for multi-step retrosynthesis planning.

ZhipuAI's **GLM-Image** (16B, MIT) converted to **Core AI** for on-device image generation on Apple Silicon (macOS 27+) — the zoo's first **autoregressive + diffusion hybrid**.

`nvidia/diar_streaming_sortformer_4spk-v2` (cc-by-4.

**The model zoo's first diffusion language model (dLLM) for Apple Core AI.

Lightricks/LTX-Video, config `ltxv-2b-0.

Hcompany/Holo2-4B converted to Apple **Core AI** for on-device inference, served by the **CoreAIChat** app.

SAM 2.

Hub catalog is a compressed-tensors (`pack-quantized`) checkpoint.

Vision-language model fine-tuned for structured data extraction from Indian financial documents.