MiniMax M2.7 vs OpenMythos

MiniMax M2.7 is a 230B-parameter Mixture-of-Experts reasoning model released as open weights in April 2026. Only 10 billion parameters activate per token (8 of 256 experts), which enables frontier-level performance at significantly lower inference cost and latency than dense models of comparable quality. The context window stretches to 204,800 tokens — roughly 307 pages of text — with strong performance on long-horizon agentic tasks. M2.7 is purpose-built for tool-using agents and coding workflows. It scored 50 on the Artificial Analysis Intelligence Index, placing it among the top open-weight models globally. Weights landed on Hugging Face simultaneously with an API launch and the open-sourcing of OpenRoom, MiniMax's interactive agent orchestration system — a rare move that gives developers the full stack from model to agent runtime. MiniMax is a Shanghai-based AI company that has been quietly iterating through M1, M2, M2.5, and now M2.7 with consistent improvements. The M2.7 release represents a notable capability jump in the MoE open-weights space, particularly for developers who need a locally deployable model that can handle complex multi-step agent tasks without calling a paid API.

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.