@fahdmirza: We trained a 9B model on Claude Mythos traces — here's what happened Meet Qwythos 9B: open-source, local, and punching …
Summary
Trained a 9B model (Qwythos 9B) on Claude Mythos traces, achieving strong results in bug-finding, SQL, and function calling while running locally on an A6000 with llama.cpp.
View Cached Full Text
Cached at: 06/28/26, 03:59 AM
We trained a 9B model on Claude Mythos traces — here’s what happened
Meet Qwythos 9B: open-source, local, and punching way above its weight class
Served locally on an A6000 with llama.cpp + MTP draft speculation Found 3 of 4 bugs in a broken full-stack call center app — one shot Nailed the SQL test: correlated subquery decorrelation, correct answer, no hints 1M context via YaRN, native function calling, fully uncensored This is what happens when small models learn from the best
Watch the full video below
Similar Articles
empero-ai/Qwythos-9B-Claude-Mythos-5-1M-GGUF
Empero AI releases Qwythos-9B-Claude-Mythos-5-1M-GGUF, a 9B parameter reasoning model fine-tuned on 500M+ tokens of Claude Mythos/Fable traces with chain-of-thought, achieving significant gains over Qwen3.5-9B and supporting 1M-token context via YaRN rope-scaling. The GGUF quantizations enable local inference on llama.cpp and compatible runtimes.
@Alacritic_Super: Need a powerful local LLM without huge hardware? Meet Qwythos-9B. • 9B parameters • 1M-token context • Native tool call…
Qwythos-9B is a 9B-parameter local LLM with 1M-token context and native tool calling, available in GGUF format for Ollama, llama.cpp, and LM Studio. It significantly outperforms its base model Qwen3.5-9B on key benchmarks.
@liu8in: Been testing Mythos for 2 hours - best code-to-motion LLM by far Claude Code + Fable 5 one-shotted this @HyperFrames_ l…
Claude Fable 5, a Mythos-class model, is announced with capabilities exceeding all previous generally available models, tested in code-to-motion tasks.
We have Mythos at Home: GLM 5.2 beats Claude in our Cyber Benchmarks
An open-weight model, GLM 5.2 from Zhipu AI, beats Claude Code in IDOR detection benchmarks at a fraction of the cost, though it still trails Semgrep's purpose-built multimodal pipeline. The article explores how much of vulnerability-detection performance comes from the model versus the harness around it.
empero-ai/Qwythos-9B-Claude-Mythos-5-1M
Empero AI releases Qwythos-9B, a fine-tuned reasoning model with 1M token context and uncensored capabilities, showing large benchmark improvements over its Qwen3.5-9B base.