RichWoollcott/studytutor-gcse-26b-moe

VISIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:2Model Size:26BQuant:FP8Context Size:32kTool Calling:SupportedPublished:May 17, 2026License:gemmaArchitecture:Transformer0.0K Featherless Exclusive Cold

RichWoollcott/studytutor-gcse-26b-moe is a 26 billion parameter Mixture-of-Experts (MoE) model, fine-tuned from unsloth/gemma-4-26b-a4b-it, designed to emulate a UK GCSE English Literature & Language tutor. This model specializes in providing explicit reasoning, modeling AQA/Edexcel marking conventions, and offering constructive feedback on student essays. With a context length of 32768 tokens, it is optimized for detailed analysis and interactive tutoring for GCSE English students.

Loading preview...

Overview

This model, RichWoollcott/studytutor-gcse-26b-moe, is a 26 billion parameter Mixture-of-Experts (MoE) model, fine-tuned from unsloth/gemma-4-26b-a4b-it. It is specifically designed to function as a UK GCSE English Literature & Language tutor, providing detailed explanations and feedback. The model was fine-tuned using Unsloth + TRL on a single NVIDIA DGX Spark GB10, utilizing multi-turn tutoring dialogues generated by an agentic-dataset-factory Player-Coach pipeline.

Key Capabilities

  • GCSE English Tutoring: Emulates a tutor for UK GCSE English Literature and Language.
  • Explicit Reasoning: Opens every response with a <think>...</think> reasoning block before the visible answer.
  • Marking Conventions: Models AQA/Edexcel marking conventions in its feedback.
  • Constructive Feedback: Provides formative feedback on student essays and writing.
  • Literary Analysis: Explains literary techniques, walks through poem/extract analysis, and models PEEL/PETAL paragraphs.

Intended Use Cases

This model is ideal as a study companion for GCSE English students, helping them with:

  • Understanding literary techniques.
  • Analyzing poems and text extracts.
  • Structuring essays using PEEL/PETAL paragraphs.
  • Receiving formative feedback on their written work.

Limitations

While the tutor persona is consistent, the model does not have access to specific exam-board mark schemes at inference time, suggesting a RAG layer for high-stakes settings. It may also hallucinate quotes from set texts, requiring verification of direct quotations.

Finetunes
1 models