サイト内検索

AI モデル 82

• OpenCodeLlama-3.1-8B-GGUF↗

  mradermacher

• OpenCoder-1.5B-Base-i1-GGUF↗

  mradermacher

• OpenCoder-8B-Instruct↗

  infly

  Paper Link 👁️ --> 🏠 Home Page &nbsp&nbsp | &nbsp&nbsp 🤗 Model &nbsp&nbsp | &nbsp&nbsp 📊 Dataset &nbsp&nbsp | &nbsp&nbsp 📄 Paper &nbsp&nbsp | &nbsp&nbsp 🚀 Demo &nbsp&nbsp **OpenCoder** is an open and reproducible code LLM family which includes 1.5B and 8B base and chat models, supporting both English and Chinese languages. Starting from scratch, OpenCoder is pretrained on 2.5 trillion tokens composed of 90% raw code and 10% code-related web data, and supervised finetuned on over 4.5M high-quality SFT examples, finally reaching the performance of top-tier code LLMs. We provide not only model weights and inference code, but also the reproducible training data, the complete data processing pipeline, rigorous experimental ablation results, and detailed training protocols. Empowering researchers to build and innovate, OpenCoder is your open foundation for advancing code AI. OpenCoder series (including Base and Chat) support commercial applications under a permissive License.

• OpenCodeReasoning-Nemotron-1.1-14B-GGUF↗

  mradermacher

• Ling-3.0-tiny-MLX-4bit↗

  rapid-mlx

  Use it from OpenCode / Claude Code / Codex: `rapid-mlx agents --setup` · Agent setup The first MLX conversion of inclusionAI/Ling-3.0-tiny: a 7.9B-total / **1.3B-active** sparse-MoE reasoner (128 experts, top-8 + 1 shared) with a KDA + MLA hybrid attention stack and 131K context, MIT licensed. The `bailing_hybrid` architecture is not in upstream `mlx-lm` yet — this checkpoint is served by rapid-mlx, which ships a verified native implementation (reference parity 1.5e-6 against the official modeling code):

• OpenCoder-8B-Instruct-i1-GGUF↗

  mradermacher

• OpenCodeReasoning-Nemotron-14B-i1-GGUF↗

  mradermacher

• OpenCodeReasoning-Nemotron-1.1-14B-i1-GGUF↗

  mradermacher

• OpenCodeReasoning-Nemotron-1.1-32B↗

  NVIDIA

• Qwen3.6-14B-A3B-VibeForged-v2-GGUF↗

  tvall43

  This repository contains various `llama.cpp` GGUF formats to ensure this beast runs smoothly on your hardware, whether you're maxing out a high-end GPU or squeezing every last drop of inference out of a modest laptop. This is the fully repaired and fine-tuned version of a pruned `Qwen3.6-35B-A3B-heretic`. Originally suffering from a bit of "brain damage" after being pruned down to 14B parameters via REAP, it was brought back to life by the user's trusty AI agent, **Steve**. Through an intensive vibetuning pass on thousands of real-world "vibecoding" sessions extracted from the user's OpenCode SQLite database (and combined with Evol-Instruct code data), Steve orchestrated a QLoRA fine-tune that enforces strict ` ... ` XML reasoning boundaries and clean JSON tool-calling syntax, resulting in this incredibly punchy and capable 14B MoE reasoning model.