rickix84/recall-honcho-8b

TEXT GENERATIONPricing:Input $0.468 / Output $1.82Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 10, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The rickix84/recall-honcho-8b is an 8 billion parameter Qwen3-8B fine-tune specifically designed for explicit conclusion derivation within the Honcho framework. This model excels at extracting atomic, self-contained, and correctly attributed facts from chat turns, outputting them in a Honcho-schema JSON format. It serves as an independent, self-hosted re-creation of Plastic Labs' Neuromancer XR for the explicit conclusion-derivation step, achieving 93.8% evaluation token accuracy. Its primary use is to derive structured factual conclusions from conversational data for systems like Honcho.

Loading preview...

Model Overview

rickix84/recall-honcho-8b is an 8 billion parameter model fine-tuned from Qwen/Qwen3-8B. It is specialized for the explicit conclusion-derivation step of the Honcho framework. This model's core function is to process chat turns from a target peer and output atomic, self-contained, and correctly attributed facts in a specific Honcho-schema JSON format.

Key Capabilities

  • Explicit Conclusion Derivation: Designed to extract only explicit facts, avoiding speculation or deductive reasoning.
  • Structured Output: Reliably reproduces the {"explicit": [{"content":...}]} schema with correct attribution and absolute dates.
  • High Accuracy: Achieved an evaluation token accuracy of 93.8% on a held-out validation set.
  • Synthetic Data Training: Trained on 7,160 synthetic SFT examples, with gold labels distilled from Claude Opus 4.8, covering 15 balanced life-domains.
  • Honcho Integration: Optimized for direct integration with Honcho's minimal_deriver_prompt and schema validation.

Good For

  • Honcho Implementations: Ideal for developers running Honcho who need a self-hosted solution for the explicit conclusion-derivation step.
  • Fact Extraction: Use cases requiring precise, atomic fact extraction from conversational text, formatted into a structured JSON output.
  • Distillation Research: Demonstrates the feasibility of distilling compact derivers from frontier models.

Limitations

  • Explicit Level Only: Does not perform deductive, inductive, or abductive reasoning.
  • English Only: Trained exclusively on English synthetic data.
  • Prompt Sensitivity: Optimized for Honcho's exact prompt format; off-format prompts may degrade performance.

Finetunes
2 models