Darwin-4B-David vs GLM-5.1

Darwin-4B-David is a 4.5-billion-parameter model that achieves 85.0% on GPQA Diamond — outperforming Google's Gemma-4-31B (84.3%) at roughly 1/7th the parameter count. The kicker: it required no training whatsoever. It was built in 45 minutes on a single H100 using MRI-guided DARE-TIES model merging, a novel variant of the merge-and-trim technique. The MRI-guided approach uses activation analysis to identify which parameters in each source model are most critical, then applies DARE-TIES merging only to the high-value weight regions. This avoids the catastrophic interference that usually degrades merged models. The result is a small model that inherits the strengths of multiple larger predecessors without any of the compute cost of fine-tuning. For the AI community, this is a meaningful data point: model merging continues to close the gap with expensive training runs. Darwin-4B-David demonstrates that thoughtful merge strategies can extract benchmark-level performance from models that are a fraction of the size, making capable AI more accessible on consumer hardware.

GLM-5.1 is Z.ai's (formerly Zhipu AI) latest open-weight model — a 744-billion-parameter Mixture-of-Experts architecture with 40B active parameters that claims the #1 spot on SWE-bench Pro with a score of 58.4, beating GPT-5.4 (57.7) and Claude Opus 4.6 (57.3). It ships under the MIT license with a 200K-token context window and maximum output of 131,072 tokens. What makes GLM-5.1 geopolitically notable is its training infrastructure: every GPU in the stack is a Huawei Ascend 910B — zero Nvidia hardware involved. This is one of the first frontier-competitive models to prove that non-Western AI compute can reach the top of benchmark leaderboards. It's a post-training upgrade to GLM-5, meaning architectural choices were locked in; the performance lift came from smarter RLHF and agentic training data. For developers, the value prop is straightforward: MIT license, frontier-level coding performance, and a 200K context window. The model is optimized for multi-step agentic tasks — it breaks down complex problems, runs experiments, reads results, and iterates. Real-world quality is still being validated beyond SWE-bench, but for teams that need a commercially-deployable open-weight coding model, this is the current benchmark king.