Solve Lite Shared CoreML Runtime

This repository hosts the frozen shared CoreML asset used by Solve Lite v0.1.11.

Current product truth

  • Compact Runtime accuracy: 82.28%
  • APS: 85.8999
  • High-confidence errors (>= 90%): 6
  • Complete installed size: 50.932773 MB
  • CoreML-native and offline after installation
  • No PyTorch or Transformers runtime
  • One shared encoder with four tiny heads: Review, Topic, NLI and Financial

v0.1.11 integration protocol

Solve Lite v0.1.11 declares 18 canonical host install profiles through one registry-driven installer. All 18 profiles passed isolated installation. This is not a claim of 18/18 product compatibility: Codex, Hermes, WorkBuddy and Doubao retain prior ordinary-session evidence; all other hosts remain PARTIAL or NOT_RUN until real-host acceptance.

Every non-empty ordinary user message enters the frozen 20-scenario router and attempts exactly one percentage block plus exactly one idempotent local reward settlement. System, assistant, tool, internal, empty and retry/replay events are excluded.

Frozen asset

  • File: solve-lite-shared-encoder-runtime-v1.tar.gz
  • Immutable asset revision: 9952d75a973e8ba4c70fddb38897480d5f1e8fd8
  • Archive SHA-256: a7359f8581bdae0c850def19bf776e540806b960c1751b0512a3002a7cf11f7e
  • Installed tree SHA-256: 857a7e0d10a1b1bf0eda59f1cad479ba6b57ca75ae9903ab098b658c1adea2a8

The v0.1.11 runtime continues to pin the immutable asset revision above. This model-card update does not replace or re-upload the model binary.

Project

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support