{"id":1715,"date":"2026-06-24T04:46:51","date_gmt":"2026-06-24T04:46:51","guid":{"rendered":"https:\/\/xrpfaucet.site\/?p=1715"},"modified":"2026-06-24T04:46:51","modified_gmt":"2026-06-24T04:46:51","slug":"meet-qwable-the-free-local-model-that-thinks-like-claude-fable","status":"publish","type":"post","link":"https:\/\/xrpfaucet.site\/?p=1715","title":{"rendered":"Meet Qwable: The Free Local Model That Thinks Like Claude Fable"},"content":{"rendered":"
\n

\ud83d\udcf0 Exclusive Crypto News & Analysis:<\/strong> Stay ahead with the latest developments in the cryptocurrency and blockchain space.<\/p>\n

\ud83d\udcc8 Market Update:<\/strong> Real-time price movements, technical analysis, and trading signals.<\/p>\n

\n
\n

In brief<\/h4>\n
    \n
  • Qwable 27B is a full fine-tune of Alibaba’s Qwen3.6-27B, trained on a Fable 5-style reasoning dataset, designed to replicate the structured, deliberate thinking style of Anthropic’s newest flagship model.<\/li>\n
  • The abliterated version removes the model’s built-in refusal behavior by surgically modifying its weights using llama.cpp’s cvector-generator.<\/li>\n
  • Both models run locally, cost nothing per query, and require neither Anthropic’s API nor its mandatory policies.<\/li>\n<\/ul>\n<\/div>\n

    Anthropic spent last week apologizing for Fable 5’s<\/a> invisible safeguards, and then the U.S. government ordered the model pulled for all foreign nationals over a disputed jailbreak finding.<\/p>\n

    A few days later, a developer on Hugging Face uploaded a model that used Fable\u2019s reasoning to guide a local model\u2014and now even your potato PC can run a better model.<\/p>\n

    The model is called Qwable<\/a>\u2014Qwen + Fable, if the portmanteau wasn’t immediately obvious. It’s a full fine-tune of Alibaba’s Qwen3.6-27B base, built by developer Mia (Mia-AiLab on Hugging Face) on a dataset of Fable 5-style reasoning examples. The goal is a 27-billion parameter model that runs on consumer hardware and thinks the way Fable 5 thinks. (Parameters determine a model\u2019s breadth of knowledge, with more generally meaning more capable.)<\/p>\n

    \n
    \n

    So I did a thing.<\/p>\n

    I have trained Qwen 3.6 27b with Fable 5 reasoning.
    Results are… interesting.<\/p>\n

    I will compare both of them side by side.<\/p>\n

    Would anyone be interesting in testing it? I can upload a gguf in hf. pic.twitter.com\/hQCiUlT1sr<\/a><\/p>\n

    \u2014 Mia (@MiaAI_lab) June 15, 2026<\/a><\/p>\n<\/blockquote>\n<\/div>\n

    The technique is called instruction fine-tuning on trace-style examples. That\u2019s a technical way of saying the developer collected examples formatted like Fable 5’s deliberate, step-by-step answers and trained Qwen to produce the same kind of output.<\/p>\n

    So think of it as less “copying the test” and more “learning the study habits.” A similar approach drove Qwopus<\/a>\u2014the Claude Opus 4.6 local distillation\u2014though that project focused on chain-of-thought reasoning traces. Qwable targets Fable 5’s overall instruction-following structure: more guided, more explanatory, and more oriented toward step-by-step task completion than the base Qwen model it was built on.<\/p>\n