AI tool comparison
Darwin-4B-David vs MiniMax M2.7
Which one should you ship with? Here is the side-by-side panel verdict, pricing read, reviewer split, and community vote comparison.
AI Models
Darwin-4B-David
4.5B merged model beats Gemma-4-31B on GPQA — no training needed
75%
Panel ship
—
Community
Paid
Entry
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.
AI Models
MiniMax M2.7
The open-source AI that improves its own training
75%
Panel ship
—
Community
Paid
Entry
MiniMax M2.7 is a 230B-parameter Mixture-of-Experts model (10B active) that does something no major open-source model has done before: it participates in its own development cycle. During training, M2.7 updated its own memory, built skills for RL experiments, and improved its own learning process — with an internal version autonomously optimizing a programming scaffold over 100+ rounds to achieve a 30% performance improvement. On benchmarks, M2.7 scores 56.22% on SWE-Pro and 57.0% on TerminalBench 2, putting it in the same tier as GPT-5.3 for coding tasks. It achieves an ELO of 1495 on GDPval-AA (highest among open-source models) and 97% skill adherence across 40+ complex, multi-thousand-token skills. For office productivity tasks — generating Word, Excel, and PowerPoint files, running financial analysis — it performs at junior analyst level. Released under MIT license on April 12, 2026, M2.7 is available on Hugging Face and via the MiniMax API. The model is particularly strong at agentic workflows: tool calling, multi-step task execution, and professional productivity use cases that require sustained context and precise instruction following.
Reviewer scorecard
“45 minutes on a single H100 to beat a 31B parameter model? That's an extraordinary efficiency ratio. MRI-guided merging is a technique I'll be watching closely. If this holds up across more benchmarks, it fundamentally changes how teams should think about building capable small models.”
“MIT license, 10B active params, and SWE-Pro scores matching GPT-5.3? This is the open-source agentic backbone I've been waiting for. The self-improvement angle is genuinely unprecedented — watching a model optimize its own scaffold over 100 rounds is the kind of thing that used to be sci-fi.”
“GPQA Diamond is one benchmark. One. Benchmark performance doesn't translate linearly to real-world task performance, especially for a merged model that hasn't been fine-tuned for instruction following or RLHF alignment. Impressive number, but I'd want to see this on coding, reasoning chains, and RAG tasks before getting excited.”
“230B total parameters is not something most people can run locally — you need serious cluster access or you're using their API, which means the 'open source' framing is mostly PR. And 'self-evolving' sounds revolutionary but the actual mechanism is AutoML loop, something the field has had for years.”
“Model merging is the dark horse of AI efficiency research. If MRI-guided DARE-TIES merging can reliably produce results like this, it suggests we're nowhere near the ceiling for extracting value from existing open-weight models. The future may involve less training and more intelligent composition.”
“A model that improves its own training process is a meaningful step toward recursive self-improvement. Even if the current implementation is narrow, this is the architectural direction that matters. MiniMax just showed a credible open-source path to it.”
“A capable model in the 4-5B range that can run on a MacBook M-series is exactly what solo creators need for on-device inference. If Darwin-4B-David's performance holds on creative tasks, it's a genuine local creative AI for people without cloud budgets.”
“97% skill adherence across 2,000-token skills means M2.7 can actually execute complex creative briefs without drifting. For long-form content workflows that need consistent style and structure, this is a real upgrade over models that forget instructions halfway through.”
Weekly AI Tool Verdicts
Get the next comparison in your inbox
New AI tools ship daily. We compare them before you waste an afternoon.