Question: where would a flash-resident tiny language runtime fit relative to on-device GenAI runtimes?

[Alpha-Guardian]

Hi ONNX Runtime GenAI folks,

I wanted to share a small on-device language-runtime experiment and ask how systems like this should be viewed relative to more familiar on-device GenAI runtimes.

We built a public demo line called Engram and deployed it on a commodity ESP32-C3.

Current public numbers:

• Host-side benchmark capability

  • LogiQA = 0.392523
  • IFEval = 0.780037

• Published board proof

  • LogiQA 642 = 249 / 642 = 0.3878504672897196
  • host_full_match = 642 / 642
  • runtime artifact size = 1,380,771 bytes

Important scope note:

This is not presented as unrestricted open-input native LLM generation on MCU.

The board-side path is closer to a flash-resident, table-driven runtime with:

• packed token weights
• hashed lookup structures
• fixed compiled probe batches
• streaming fold / checksum style execution over precompiled structures

So this is not a standard portable GenAI runtime story. It is closer to a task-specialized language runtime whose behavior has been compiled into a very constrained execution form.

Repo:
https://github.com/Alpha-Guardian/Engram

What I’m curious about is whether systems like this should be viewed as:

• an extreme on-device specialization endpoint
• outside the normal GenAI runtime family
• or an adjacent future category for highly constrained language systems

Would love to hear any thoughts.