For Immediate Release
TuitionGoWhere Starts Model-Labelled Tencent HY3 Generation Trial for Quiz and Exam Paper Variants
The July 2026 run added a new OpenRouter-hosted model variant to TuitionGoWhere's generated quiz and exam paper pipeline and was followed by a separate benchmark review.
SINGAPORE, July 2026 - TuitionGoWhere has started a new model-labelled generation trial using Tencent HY3 Free through OpenRouter to create additional quiz and exam paper variants for Singapore students.
The trial is part of TuitionGoWhere's ongoing effort to compare how different AI models perform when generating free test papers, free exam papers, and topic quizzes across Primary, Secondary, O-Level, and A-Level pages. The HY3 run is configured as a separate model suffix so its output can be reviewed independently from existing Qwen, DeepSeek, Gemma, Kimi, Owl, Nemotron, and Claude-labelled resources.
The generation process uses two concurrent workers and follows the same modular pipeline used by other model runs. Where exam-derived templates are available, the generated material is asked to follow the structure and style of past-paper question patterns. Where such evidence is limited, the pipeline falls back to syllabus-backed generated templates and labels the output as AI-generated practice material rather than exam-derived content.
No Official MOU or Partnership
This announcement is not an official memorandum of understanding, partnership, endorsement, procurement arrangement, or collaboration agreement with Tencent or OpenRouter. It is a technical generation run using a publicly listed OpenRouter model endpoint, with model provenance preserved for later quality review.
Benchmark Review
The completed review covered 2,107 HY3-labelled quiz and paper reports using the scoring framework shown on the platform's benchmark dashboard. It examined language fit, syllabus alignment, question-template discipline, artefact cleanliness, answer explanation quality, notation handling, paper format, difficulty, and whether papers were realistically doable within the stated timeframe. The results are published in the Tencent HY3 Free Benchmark Whitepaper.
Why Model-Labelled Variants Matter
Model-labelled variants make it easier to compare output quality without hiding the source model. This allows the platform to keep stronger resources visible, identify weaker cases for regeneration, and improve future prompts without treating every AI output as equally reliable.
About TuitionGoWhere
TuitionGoWhere is a Singapore-focused education project offering free AI-powered practice papers, topic quizzes, and learning resources. It is built and maintained by IT professionals as a community contribution.