AI tool comparison
OpenMythos vs Talkie
Which one should you ship with? Here is the side-by-side panel verdict, pricing read, reviewer split, and community vote comparison.
Research & Open Source
OpenMythos
Open-source PyTorch reconstruction of Claude Mythos' suspected architecture
75%
Panel ship
—
Community
Paid
Entry
OpenMythos is a PyTorch reconstruction of the suspected architecture underlying Anthropic's Claude Mythos model, built entirely from published research. Creator Kye Gomez hypothesizes that Mythos uses a Recurrent-Depth Transformer (RDT) — where a subset of transformer layers loops multiple times per forward pass with shared weights rather than stacking unique layers. This allows the model to simulate "thinking" by iterating over the same compute graph, giving it emergent chain-of-thought behavior without explicit CoT prompting. At 770M parameters, the OpenMythos implementation reportedly matches the downstream quality of a 1.3B standard transformer on benchmarks. The architecture combines Multi-Latent Attention for memory compression, LTI (Linear Time-Invariant) stability constraints to prevent training instability during recurrence, Mixture of Experts routing for specialization, and Adaptive Computation Time (ACT) halting to decide when to stop looping per token. The project exploded on GitHub within days — 6.2k stars, 1.2k forks — and Kye's X announcement drove massive engagement (4.1k likes, 4.5k reposts). Community reaction is genuinely divided: AI researchers calling it "the most sophisticated reverse-engineering of an LLM architecture I've seen" while Anthropic has not confirmed or denied any of the architectural claims. This is an educated speculation backed by real engineering, not a marketing exercise.
Research
Talkie
A 13B LLM trained exclusively on texts from before 1931
75%
Panel ship
—
Community
Free
Entry
Talkie is a 13-billion parameter language model trained exclusively on English-language texts published before 1931 — the largest vintage language model built to date. Created by researchers Nick Levine, David Duvenaud (University of Toronto), and Alec Radford (of GPT and DALL-E fame), it represents a novel approach to understanding what training data really does to a model. The research insight is elegant: modern LLMs are so thoroughly contaminated by modern internet data (directly or through distillation) that it's nearly impossible to isolate what the model "knows" from what it absorbed during training. Talkie solves this by hard-cutting the training corpus at 1931 — predating digital computers entirely. This lets the team run controlled experiments impossible with contemporary models, such as teaching the model to write Python from examples alone and measuring how quickly it generalizes. Talkie was trained on ~260 billion tokens of historical text and fine-tuned using direct preference optimization with Claude as judge on structured historical documents (etiquette manuals, letter-writing guides). It's openly available on Hugging Face for research use. It also happens to produce wonderfully formal, slightly anachronistic prose.
Reviewer scorecard
“Whether or not Anthropic actually uses this architecture, the RDT implementation itself is genuinely impressive engineering. The ACT halting mechanism and LTI stability constraints are clever solutions to problems anyone trying to build reasoning models will face. Fork-worthy regardless of the Mythos speculation.”
“The ability to test code-learning from scratch on a model that's never seen a modern codebase is genuinely useful for ML research. The methodology here is cleaner than anything I've seen for studying data contamination.”
“This is reverse engineering based on vibes and published papers, not leaked weights or verified architecture docs. Anthropic hasn't confirmed a thing. The 770M benchmark comparisons are cherrypicked and the '1.3B equivalent quality' claim needs independent reproduction. Intellectually interesting, empirically unverified.”
“Fascinating as a research artifact, but this isn't a production model. The limited vocabulary and cultural frame mean it's not useful for most practical tasks. It's a museum piece, not a tool.”
“Regardless of whether Mythos actually is an RDT, this project demonstrates that open-source researchers can meaningfully reconstruct competitive reasoning architectures from scratch. That capability gap between frontier labs and open-source is closing faster than most realize.”
“This is exactly the kind of fundamental research the field needs. Understanding what training data does to language models — not just benchmark scores — is critical as we scale to more powerful systems. Radford's involvement adds serious credibility.”
“A 6.2k star project in two days means something hit a nerve. The documentation is excellent — clear architecture diagrams, detailed training notes, working code. Even if the Mythos speculation is wrong, this is a model for how to share research engineering properly.”
“The prose it generates has a formal, unhurried quality that modern LLMs can't replicate. For period-accurate creative writing, historical fiction, or vintage-voice content, Talkie is the only model worth using.”
Weekly AI Tool Verdicts
Get the next comparison in your inbox
New AI tools ship daily. We compare them before you waste an afternoon.