Tencent HY3 Free Benchmark Whitepaper
Tencent HY3 Free produced strong language, notation, and syllabus scores, but the benchmark found weaker difficulty calibration and recurring format problems in some subjects.
Read whitepaperPress releases, benchmark whitepapers, and official announcements for TuitionGoWhere, a free AI-powered learning resource for Singapore students.
Tencent HY3 Free produced strong language, notation, and syllabus scores, but the benchmark found weaker difficulty calibration and recurring format problems in some subjects.
Read whitepaperQwen3.6 Plus had the strongest large-sample quiz and paper average in the benchmark, with good notation handling and strong syllabus fit.
Read whitepaperDeepSeek V4 Pro was one of the most stable large-sample models, with only four quiz or paper reports below 8.0.
Read whitepaperGemma 4 31B produced clean, readable quiz and paper content, but the benchmark showed a clear weakness in answer explanation depth.
Read whitepaperLongCat 2.0 LLM generated the largest quiz and paper sample in the benchmark and produced many usable resources, but it also had more low-score cases.
Read whitepaperKimi K2.6 Free was broadly solid, with strong notation scores and good subject coverage, but it needs tighter checks for answer steps and difficulty.
Read whitepaperClaude Sonnet 4 was strong for clean prose, parent guides, cheatsheets, and audio scripts, while quiz and paper templates needed stricter control.
Read whitepaperQwen3.7 Plus had excellent notation handling but a higher low-score rate in the smaller rerun sample, especially around difficulty and language control.
Read whitepaperMiniMax M3 had the highest average score, but the sample was only 29 Primary 1 reports, so the result should be treated as early evidence.
Read whitepaperThe July 2026 generation run added a Tencent HY3 Free model-labelled variant through OpenRouter, followed by a 2,107-report benchmark review.
Read releaseThe roadmap outlines future ELI5 lessons, audio-first learning formats, and experimental supports for students with different learning needs, including dyslexia and ADHD/ADD.
Read releaseThe January 2026 launch introduced free Primary 1 to Primary 6 practice papers and learning resources aligned with Singapore MOE syllabus expectations.
Read releaseThe June 2026 expansion introduces model-labelled Secondary, O-Level, and A-Level quiz and paper generation using newer frontier models through OpenRouter.
Read releaseThe release expands Chinese, Malay, and Tamil practice across primary and secondary levels, supporting Singapore families who want Mother Tongue learning to remain active beyond the classroom.
Read release