Open models
Useful ideas. Usable artifacts.
Seven public releases: a decision model and its formats, and EEG foundation models packaged for Apple platforms. Each card leads with what the model is for, how it was measured, and where it stops.
Manchegov3Manchego V3: a 4B typed-decision model. State, question and options in; probabilities out, in one forward pass.Manchego · MLX 8-bitv3Manchego V3 for Apple silicon. 8-bit MLX, 4.5 GB, with the SemIf decision contract and published format checks.Manchego · MLX 4-bitv3Manchego V3 for Apple silicon. 4-bit MLX, 2.4 GB, with the SemIf decision contract and published format checks.ZUNA1.1 · Core MLfp32 profilesZyphra’s ZUNA1.1 on Apple platforms: EEG denoising, missing-channel reconstruction, and spatial superresolution in Core ML.CBraMod · Motor imageryCore MLA cue-paced left/right motor-imagery classifier for a 14-channel EPOC X-style montage, in a single Core ML model.CBraMod · EmbeddingsCore MLThe pretrained CBraMod backbone as a native Core ML embedder. A 14-channel, five-second window in; a 14,000-dimensional embedding out.ZUNA · Core MLFive profilesThe original ZUNA reconstruction model as five enumerated Core ML profiles, from a 14-channel EPOC X layout to 64 channels.
No models found.
Try a different name, runtime, or research area.
7 models shownCurated snapshot · reviewed 23 Oct 2026
↳Provenance
Where every artifact comes from.
Upstream models, Oraculum’s work, and each published format. Upstream authors keep their credit.
Decision models
- Qwen/Qwen3.5-4BUpstream · Apache-2.0
- V3 · fresh rank-16 LoRA255,991 rows · 99.4M prompt tokens
EEG reconstruction
EEG reconstruction
EEG representation and classification
- CBraModUpstream · braindecode checkpoint
- CBraMod · EmbeddingsBackbone, head removed
- Motor-imagery fine-tuneOraculum · 109 subjects
Follow the work
Good questions. Measured answers.
New models, their evidence, and the occasional correction. Model cards come first; this site follows.