DeepSeek V4 vs OpenMythos

DeepSeek V4 dropped April 24, 2026 as two production-ready Mixture-of-Experts models: V4-Pro (1.6T parameters, 49B activated) and V4-Flash (284B parameters, 13B activated). Both support 1 million token context and ship under the MIT license — the most permissive option in AI. The architecture innovation is the hybrid attention mechanism combining Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA), which slashes long-context inference costs dramatically. At 1M tokens, V4-Pro requires only 27% of the FLOPs and 10% of the KV cache compared to DeepSeek V3.2 — a meaningful efficiency gain that makes million-token context economically viable. Performance-wise, DeepSeek V4-Pro beats all rival open models on math and coding benchmarks, trailing only Google's Gemini 3.1-Pro (closed) on world knowledge. One year after V2 upended the industry, DeepSeek has done it again — a model approaching frontier performance that anyone can run, modify, and ship commercially with zero licensing friction.

OpenMythos is a community-driven theoretical reconstruction of Claude Mythos's suspected architecture, implementing a Recurrent-Depth Transformer (RDT) — a looped transformer that recycles layers multiple times per forward pass for deeper reasoning without massive parameter growth. The project drew 10,100 GitHub stars in its first week, reflecting intense developer curiosity about what's powering Anthropic's latest generation models. The architecture has three stages: a Prelude (initial layers), a Recurrent Block (looped up to 32 times with shared weights), and a Coda (final layers). Rather than stacking hundreds of unique layers, the recurrent block runs the same weights multiple times with learned injection parameters updating hidden states between loops — enabling implicit chain-of-thought reasoning in continuous latent space without generating intermediate tokens. The project supports Grouped Query Attention (GQA) with optional Flash Attention 2, Multi-Latent Attention (MLA), and sparse MoE with routed and shared experts. Model scales range from 1B to 1T parameters. The key claim is that RDT achieves reasoning depth comparable to fixed-depth models with far more parameters, since computational complexity scales with loop iterations rather than layer count. This would explain how Claude Mythos achieves strong reasoning performance without the extreme parameter counts of brute-force scaling — though Anthropic has neither confirmed nor denied the architecture.