Havuz

网页版登录态 → OpenAI 兼容本地反代：DeepSeek / Qwen / Doubao 免费额度驱动 ZCode。Web login → OpenAI-compatible local API.

Hemmingway-1: a 27B model that writes the way a person writes. Open weights, free for non-commercial use (CC BY-NC 4.0); commercial use by a

An inference framework customized for AI MAX 395, compatible with both Linux and Windows, and supports deploying only qwen3.8-flash-next. Th

Hybrid GPU + CPU inference for mixture-of-experts models, written in Rust down to the CUDA kernels. A drop-in llama-server.

Run SemIf (Jev-style semantic-if decisions) on a CPU — no GPU. Reads typed option probabilities straight from an open model in one forward p

Open-source alternative to TypeSafe's Jev: a System One style model layer that gives typed, calibrated decisions from any open-weights LLM i

Typed decisions (choice / score / yes-no) from local Qwen models on Apple Silicon. Probabilities come straight from the logits, no text gene

Refusal removal as a loadable KV-cache graft — zero weight modification, no refusal-direction projection, fully reversible.

AI 厂商动态别挨家追：每天巡检 OpenAI、Claude、Gemini、DeepSeek、通义、Kimi 等的官方博客与更新页，生成可订阅的 RSS，英文原文逐段精译为中文，并附 70 家免费额度 / 限时活动总表（实时抓自官方页、带官方链接可复核）

BrighTO LLM Router: free open-source, ultra-fast Rust LLM gateway/router and LiteLLM/Bifrost alternative with OpenAI + Anthropic APIs, Claud

食毅：已上线的 AI 饮食记录微信小程序，采用多模态识别、Human-in-the-loop 与确定性营养护栏

An evidence-grounded LLM agent for diagnosing PyTorch training failures and verifying proposed repairs through bounded retraining.

A NPC AI Framework , based on character personalities, can be used for game NPC (extendable for Godot / Unity)

Logic-based multi-LLM voice router for Home Assistant, combining native intents, local Qwen, GPU-aware desktop Ollama, and cloud AI with int

Advanced local AI prompt generator for image & video models — reference images, dataset creation, prompt refinement, model-specific outputs

Raw benchmark logs for Swift-Qwen3.8-27B vs the Qwen3.8-27B BF16 base: every response, score, and config across 9 benchmarks

Self-hosted voice gateway in Go that speaks the OpenAI Realtime (GA) WebSocket protocol and runs a cascaded ASR → LLM → TTS pipeline behind

High-performance OpenAI & Anthropic compatible API gateway for Qwen with multi-account rotation, interactive TUI dashboard, and resilient to

迹录 Tracord —— 基于本地小模型多智能体的智能旅行规划助手 / AI travel planner with local SLM multi-agents — light up visited cities on a map, log expenses with one

Cross-model transfer of persistent recurrent inference state without prefix replay

High-performance CUDA inference engine for Qwen models, written from scratch in C++/CUDA

llama.cpp fork for significantly improved performance on Ampere generation Nvidia cards (especially RTX 3090 / 3090 Ti): example: 110+ tok/s

Ollama-compatible LLM server on a tuned llama.cpp fork. Beats vllm by 40% on single stream decode, and by 20% on multi stream decode, and be

ENGRAFT: write facts into an LLM's n-gram (Engram) memory table as a removable overlay: 100 facts in a 9 MB file, no weights touched, verifi

Can't find subtitles? Generate SRT/ASS subtitles from any video with local AI and auto-translate them — no internet, free. macOS & Windows.

Hybrid rule-based and context-aware LLM evaluation system for Chinese-to-Japanese game localization QA.

Mercury CLI: multiple AI coding sessions in one terminal. Mix and match provider and locally hosted models, spawn cross-company agents & wor

The best model under 200B at near 10 token/s exact batch throughput on a single laptop CPU, with an automatic 8 GB RAM path. Native C, no GP

《大学生AI全场景使用指南》· 13 个校园场景的千问实操手册：逐步截图、指令原文与可复用提示词模板

The open-source Three.js-based modern LLM Engine that runs in your browser via WebGPU compute.

OpenAI-compatible inference server for Qwen3.8 on one RTX 5090 / RTX PRO 6000: the 27B at ~170 tok/s (peaks ~370), Flash-Next (125B MoE) at

Run a 177B model (Qwen3.8-Flash-Next) on an 8 GB laptop GPU. Measured: 6.6 GiB VRAM, 47.8 GiB RAM, 34-35 tok/s. The n-gram table stays on di

AI-enabled Intelligent Document Processing for Python: 12+ LLM backends, Pydantic schemas, self-hosted OCR (Nanonets-OCR2-3B), auto-schema d

Open-source CLI bridge for shell-capable AI agents to work in existing Chrome tabs. Local Node server, no third-party runtime dependencies,

Slotstream, büyük bir MoE modelini SSD’den parça parça yükleyip yoğun kullanılan expert’leri bellekte önbelleğe alarak Apple Silicon Mac’lerde çalıştıran yerel bir inference engine’dir.

Neden yükseliyor · README, 48 GB M5 Pro üzerinde belirli koşullarda 15.86 tok/s ölçümü ve prompt işleme iyileştirmeleri yayımlıyor; son 28 günde 271 commit ve 5 farklı yazar da yoğun geliştirme faaliyetine işaret ediyor. Ancak proje yalnızca 30 günlük ve mevcut sürüm için nitelikli tam yanıt hızı ölçümünün hâlâ beklediği belirtiliyor.

Durable local inference for Oh My Pi: NInfer + Qwen3.8 27B on one RTX 3090/4090/5090, with restart-resumable OpenAI Responses state.

What I run on my inference server: Qwen3.8-Flash-Next on two RTX 3090s, and how I got here

halogen-flash-server, Qwen3.8-Flash-Next modelini AMD Strix Halo (gfx1151) donanımında çalıştırmak için özelleştirilmiş bir inference server'dır.

Neden yükseliyor · 32 günlük repo için 722 yıldız ve son 28 günde 99 commit dikkat çekiyor. README, belirli donanım üzerindeki ölçümleri ve ölçüm koşullarını açıklıyor; ancak performans iddiaları proje tarafından sunulmuş ve bağımsız doğrulama bilgisi burada yok. 99 commit'in tek yazardan gelmesi geliştirme hızını gösterirken bakım kapasitesine dair kanıtı sınırlıyor.

The fastest way to run Qwen3.8 27B on Strix Halo (gfx1151)

Agentic RAG consumer decision agent with ReAct, Text2SQL, multi-hop retrieval, memory and deterministic constraint checking.

模型路由器：在 Codex Desktop 和 Claude Desktop 使用 Claude、第三方 API 与自定义模型。Local LLM proxy for Codex & Claude Desktop: OpenAI Responses / Chat Completi

Serve DeepSeek and Qwen, run an on-demand model library, and fine-tune with Unsloth QLoRA on 2x NVIDIA DGX Spark. TP=2 vLLM lanes behind one

Evaan — a fully local, offline-capable AI companion chatbot built with Python & Hugging Face Transformers, running on Qwen2.5-0.5B-Instruct

Qwen3.8-27B on one laptop CPU: up to 2.52 token/s, 8 GB tested, no runtime accuracy loss. Native OpenAI-compatible API with function tools;

I Ran Qwen 3.8 27B on Just 4GB VRAM! (FreeToken Tutorial) - Edge-native MoE serving engine for local LLMs on low VRAM GPUs

One command spins up Qwen3.8-27B (abliterated, NVFP4) on a rented Blackwell GPU at ~150 tok/s and wires it into your local coding agent via

Docker compose configs for serving Qwen3.8-27B on a DGX Spark (GB10): vLLM+MTP current, SGLang+DSPARK + FP8 rollback stacks

Winning Qwen and DeepSeek setups for AMD Strix Halo: pinned Ansible recipes and matched benchmarks on a 128 GiB Ryzen AI Max+ 395 / Radeon 8

The local AI app built for the Mac. Run LLMs like Qwen, Gemma and GPT-OSS on Apple Silicon, including models far bigger than your Mac's memo

Serving Qwen3.8-27B-FP8 on a single DGX Spark (GB10): 7.88 to 58.5 tok/s single-stream from decode strategy alone, weights untouched. Specul