Everything I've released or helped release, with what I actually did on each one. Download counts come from the Hugging Face API, all-time, fetched September 25, 2026.
27B dense502,598 all-time downloads
A code-heavy finetune had broken the model’s chain of thought. Fusion takes the difference between the coder and its stable parent as a task vector and merges it back in, a little at the bottom of the network and more toward the top. Reasoning stays intact and the coding gains carry over.
18B743,796 all-time downloads
A frankenmerge: transformer layers stacked into a new 18B model, then healed with QLoRA so the seams stop showing. The frontier models I asked advised against it. It became one of my most downloaded releases, and the full recipe is published step by step.
The Qwopus line is led by Jackrong. My contributions are compute (the GB10 in my lab), data curation, tooling, and independent evaluation. The role on each row is what the model card credits.
27B dense282,043 all-time downloads
Qwen 3.8 27B tuned to think less wastefully in agent loops. SFT on the best 10% of about 1.5M teacher examples, then NeMo-RL with GSPO. We cut runaway reasoning and sped up decoding, and paid a small price on MMLU-Pro that the card publishes.
Tradeoff, published on the card: MMLU-Pro mixed set: 91.28% vs 92.73% for the base (−1.45 points).
27B dense and 35B-A3B MoE
The Qwopus 3.6 line, led by Jackrong. I contributed training compute, data curation, a tool-calling chat template that works across runtimes, agentic demo Spaces, and independent evaluations, and I’m credited on the cards for hardware infrastructure, training support, and live agent experiments.
4B trace inverter; credited on the 9B card
A 4B model that runs reasoning backwards: it expands compressed reasoning summaries into full, teachable chains of thought. Curated sets of 5k to 9k expanded traces then serve as training curriculum.
Link coming soon
Swift and Apple-platform specialist. The first release from the domain distillation workstream. Why Swift
Qwopus 3.8. The line continues on the Qwen 3.8 base, with much of its data reused from curated 3.6 trajectories.
Follow releases on Hugging Face and X. I answer nearly every question in release threads and fold the feedback into later versions.