Aurekai

Try a model

BitNet Embedding 270M I2_S

Official BitNet embedding release body used only through the distinct BitNet representation passport. This identity does not imply comparability with any Bonfyre or provider embedding geometry.

Your input

Checking whether you can run this model…

Answer

The answer appears here. Nothing runs until you press Run.

About this model

Format
gguf
Size
350 MB
Shape
640-dimensional normalized embedding
Family
none
Tested for diagnostic
measured, on hold

bitnet-embedding-270m-i2s-pr606-2026-09. Nine repeated executions across one, two, and four threads produced the same finite 640-dimensional candidate vector on the CPU-only x86 compatibility runtime. One real AddressPlane run deterministically reduced 4891 external occurrences to 128 mask survivors and reordered all 128 in SHADOW state. A separate complete native ForeignTwin retained-body evaluation recorded 104 retained observations across that 128-member frontier, including 15 retained at rank 16 against an 81.25 percent frontier baseline; this is one owner predicate on one query, not retrieval preservation, semantic relevance, top-k recall, false-prune safety, temporal stability, cross-model comparability, or a semantic quality envelope. The representation remains diagnostic-only and may reorder a bounded AddressPlane survivor set in SHADOW mode; it may not truncate, select, authorize, verify, lower, or enact. Check code e33c4afde347c522b90cbaf2560255130a3bc9ba2129cd93eec63ce7b105c255.

Price
No price published

No provider lists a price for this model here. That doesn’t make it free to run.

Use it in your work

Being listed here doesn’t mean you can run it. Running needs the model installed, access for your account and a limit you check before it starts.