Bonsai-8B vs OpenMythos

Bonsai-8B is a 1-bit quantized language model from Prism ML, based on Qwen3-8B, that compresses a full 8B parameter model down to just 1.15 gigabytes. Running at 368 tokens per second on an RTX 4090, it achieves a 6.2x throughput speedup over FP16 equivalents while scoring 70.5 average across standard benchmarks — maintaining competitive quality despite the extreme compression. The model uses end-to-end 1-bit quantization rather than post-training quantization applied to a pretrained FP16 model. This means all weights are trained natively as ternary values {-1, 0, +1}, enabling the 14x size reduction versus FP16 without the quality cliff typical of aggressive post-training quants. Bonsai-8B targets the edge and on-device inference market: robotics, mobile apps, offline-capable applications, and scenarios where privacy and latency requirements make cloud inference impractical. The 1.15GB size fits in phone RAM and runs on consumer CPUs. Apache 2.0 license means it's deployable anywhere.

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.