Pending benchmarksTracking since Nov 4, 2026
Is Pyramid-JiT 2B (pjit-21k) state of the art?Pending benchmarks.Announced Nov 4, 2026; no independent results yet.
Linum's 2.2B pixel-space text-to-image research checkpoint trained for about $21K, with open weights and code.
Announced facts
The details so far
- Lab
- Linum
- Announced
- Nov 4, 2026
- API
- Not yet
- Open weights
- Published
- Parameters
- 2.18B (2,184,091,606)
- Training cost
- About $21K (lab-stated)
- Training samples
- 660M
- Text encoder
- Frozen Qwen3.5-4B
- License
- Apache-2.0 (weights and code)
- Limitations
- Research checkpoint; cannot render legible text
- Lineage
- Uses the P-JiT pixel-space architecture Linum published in October 2026
Tags marked unsure are not stated on the lab's own pages.
Next steps
What would let us call it
Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.
- Publicly announced with a source
- Usable by the public (API or weights)
- Listed on an independent board we track
- Enough results for a rank in a category
- Holds across two snapshots (crowns only)
Receipts
Sources
LABTraining a Text-to-Image Model for $21Klinum.ai/field-notes/pjit-21k · Nov 4, 2026REPOLinum-AI/pjit-21k on Hugging Facehuggingface.co/Linum-AI/pjit-21kREPOLinum-AI/pjit-21k on GitHubgithub.com/Linum-AI/pjit-21kTracking continues until an independent board lists it.