Skip to content

Repository files navigation

OpenMythos-MLX 🍎🐦‍⬛

OpenMythos ported to Apple Silicon MLX — with CONFIRMED 4x depth extrapolation!

MLX License: MIT

Disclaimer: OpenMythos is an independent, community-driven theoretical reconstruction. Not affiliated with Anthropic.

🔥 BREAKTHROUGH: 4x Depth Extrapolation Confirmed

Trained on RavenX-Sec security data, M4 Max 128GB:

n_loops= 1: loss=10.3155
n_loops= 2: loss=10.2770  ← trained here  
n_loops= 4: loss=10.2448  ← BETTER (extrapolated!)
n_loops= 8: loss=10.2380  ← BEST (4x training depth!)
n_loops=16: loss=10.2488  ← slight degradation
n_loops=32: loss=10.2644  ← ACT halting would fix this

Train at 2 loops → optimal at 8 loops = 4x depth extrapolation. More loops = deeper reasoning without additional parameters.

This is the first demonstration of RDT depth extrapolation on Apple Silicon and on security domain data.

What This Is

A pure MLX implementation of the OpenMythos architecture — a Recurrent-Depth Transformer (RDT) that uses the same weights looped T times for deeper reasoning without parameter growth. Now includes MoDA (Mixture-of-Depths Attention) where each layer attends to ALL preceding layers' outputs.

Original: github.com/DeadByDawn101/OpenMythos (PyTorch)

Two Architectures

RDT (Recurrent-Depth Transformer) — model.py

Input → [Prelude] → [Recurrent Block × T loops] → [Coda] → Output
                      ↑_____________↓
                      Same weights, more loops = deeper thinking

MoDA (Mixture-of-Depths Attention) — moda_mlx.py

Layer N attends to: [current sequence] + [Layer 1..N-1 outputs]
= The model can SEE ITS OWN THINKING from previous layers!

Validated by Anthropic's Claude Mythos Architecture

On June 9, 2026, Anthropic released the Claude Mythos 5 / Fable 5 System Card and the Claude Mythos Preview System Card. Their findings validate EVERY core concept in OpenMythos.

We were already on the same path — building independently on Apple Silicon what Anthropic built at $100M+ scale.

OpenMythos Concept Anthropic's Parallel System Card Section
Depth Extrapolation (train 2 → test 8) "Substantial post-training and fine-tuning" — iterative refinement across multiple model snapshots §1.1, §1.1.4
Thinking Toggle (OFF/LOW/MED/HIGH) "Chain-of-thought controllability" — testing reasoning trace control and monitorability §6.5.5 (Mythos 5)
Multi-Trajectory Reasoning (GRAM best-of-N) Multi-Agent Harnesses — exploring multiple reasoning paths simultaneously §8.15 (Mythos 5)
Progressive Training Rounds (12 rounds) "Different snapshots taken at various points during training" with iterative evaluation §1.4 (both cards)
Builder/Breaker (OpenSelfRevise) Adversarial safety testing — "destructive or reckless actions in pursuit of goals" §4.3.1 / §6.3.1
Safety at Model + Harness Level Classifier fallback architecture: Fable 5 falls back to Opus 4.8 when classifiers trigger §1.5 (Mythos 5)
Thinking Block Stripping ( removal) "Reasoning text is denser and more difficult to interpret" — monitoring and controlling thought traces §6.5.5.1 (Mythos 5)
Agent Memory + Provenance "Automated offline monitoring" of behavior patterns across sessions §4.2.1.2 / §6.2.1.2
Cyber as Top Domain "Most capable model on cyber tasks" — "autonomously discover and exploit zero-day vulnerabilities" §3 (both cards)
Dynamic Profiles (mode switching) Safety classifier routing — different capability levels per task type §1.5 (Mythos 5)

Key Quotes from the System Cards

On iterative refinement: "Different snapshots of the model are taken at various points during the training process" — exactly our progressive round training methodology.

On cyber capabilities: "Mythos 5 is the most capable model we have evaluated on cyber tasks... scores far ahead of Claude Opus 4.8" — our RavenX-CyberAgent operates in this same space.

On reasoning control: "Chain-of-thought monitorability evaluations" — validates our thinking toggle and <think> block management.

On safety architecture: "Fable 5's cybersecurity classifiers are effective at detecting cyber use and cause the model to fall back to Opus 4.8" — same concept as our Dynamic Profiles and harness-level safety.

On model awareness: "Evaluation awareness is significant and not always verbalized" — critical insight for our Evaluations framework design.

What This Means

We built the open-source version of the same methodology Anthropic uses for their most powerful model.

ANTHROPIC (closed, $100M+):          OPENMYTHOS (open, 1 MacBook):
Iterative post-training               → Progressive rounds (12)
Chain-of-thought control               → Thinking toggle (OFF→HIGH)
Multi-agent harnesses                  → GRAM multi-trajectory
Classifier-based fallback              → Dynamic Profiles
White-box activation analysis          → Core AI Debugger (Apple)
Adversarial safety testing             → Builder/Breaker (OpenSelfRevise)
Cyber as top capability domain         → RavenX-CyberAgent
Safety at model + system level         → Dual-level memory (model + harness)

The name isn't a coincidence. The methodology isn't a coincidence. We were on the same path.


Quick Start

import mlx.core as mx
from open_mythos_mlx import OpenMythos, mythos_1b

cfg = mythos_1b()
model = OpenMythos(cfg)

ids = mx.random.randint(0, cfg.vocab_size, (1, 32))
logits = model(ids, n_loops=8)  # more loops = deeper reasoning

Key Innovation: stop_gradient Training

The breakthrough for stable RDT training on MLX:

for t in range(n_loops):
    if t < n_loops - 1:
        x = mx.stop_gradient(x)  # detach history
    out = self.recurrent(x + 0.1 * e)
    x = 0.5 * x + 0.5 * out

Only backpropagate through the last loop iteration — prevents gradient explosion through time while maintaining depth extrapolation.

Research Papers This Enables

  1. "RDT-Distilled Security Reasoning in MoE Transformers" — Train RDT on security data, distill deep traces into production MoE model
  2. "RDT-to-MoE Reasoning Transfer" — General technique for transferring variable-depth reasoning to fixed-depth architectures

Credits

License

MIT

Sponsors

OpenMythos-MLX Sponsors

Become a sponsor on Tribe.run.

About

OpenMythos ported to Apple Silicon MLX — Recurrent-Depth Transformer with MoE, MLA, ACT halting

Resources

Stars

15 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages