Quiz AI Report Card
We gave free AI models the same 10 grades 3–8 quiz tasks, from 3rd-grade fractions to 8th-grade forces and motion, checked every answer key question by question, and scored each output 1–5 on answer-key accuracy, grade fit, classroom-readiness, question quality and format. Every output is published in full, so you can check our grading.
Tested 2026-10-164 models · 10 tasks · 40 scored outputsHow we score →
Biggest finding: every model wrote at least one wrong answer key. Qwen3.8 27B (4.14) and GPT-OSS 20B (4.12) are essentially tied at the top. GPT-OSS 120B, the largest model, came third (3.80) because it mis-keyed 8.4 ÷ 0.2 as 16.8 (it's 42), called a non-proportional table proportional, said New England had three colonies (it had four), and left "Wait, check" self-corrections in a key. ALLaM 2 7B (1.58) is not usable for quizzes. Its keys rarely match its questions.
Check before you print: figurative-language keys were weak for every model (for example, "The classroom was a zoo" keyed as personification, and "The cat sat on the mat" keyed as alliteration). Several quizzes had two correct choices or bolded the right answer in the student copy. Science (cells, forces) and 8th-grade equations were the most reliable.
Note on Qwen3.8 27B: Groq's free tier limits it to 1,000 output tokens per minute, so all 10 of its quizzes ran with a 1,500-token budget (the other models got 3,000). Three of its answer keys were cut off (math5, math7, sci8) and were marked down for it.
Rankings
| # | Model | Overall | Best area | Free tier | Privacy | |
|---|---|---|---|---|---|---|
| 1 | Qwen3.8 27B Qwen (Alibaba) · open model via Groq | 4.1/5 | Social studies (4.8) | Groq free tier: 30 requests/min, 1,000 requests/day, 8K tokens/min. source | Groq says it does not retain inference inputs/outputs by default; it may log them for up to 30 days for reliability or abuse checks. Never paste student names or personal info. source | Scores & full outputs → |
| 2 | GPT-OSS 20B OpenAI (open-weight) · open model via Groq | 4.1/5 | Science (4.4) | Groq free tier: 30 requests/min, 1,000 requests/day, 8K tokens/min, 200K tokens/day. source | Groq says it does not retain inference inputs/outputs by default; it may log them for up to 30 days for reliability or abuse checks. Never paste student names or personal info. source | Scores & full outputs → |
| 3 | GPT-OSS 120B OpenAI (open-weight) · open model via Groq | 3.8/5 | Science (4.3) | Groq free tier: 30 requests/min, 1,000 requests/day, 8K tokens/min, 200K tokens/day. source | Groq says it does not retain inference inputs/outputs by default; it may log them for up to 30 days for reliability or abuse checks. Never paste student names or personal info. source | Scores & full outputs → |
| 4 | ALLaM 2 7B SDAIA · open model via Groq | 1.6/5 | Social studies (2.0) | Ran on our Groq free-tier account; Groq's published limits table does not list this model. source | Groq says it does not retain inference inputs/outputs by default; it may log them for up to 30 days for reliability or abuse checks. Never paste student names or personal info. source | Scores & full outputs → |
Open models were tested through the free Groq API (one run per task, temperature 0.3). Every score is one run per task: a snapshot, not a verdict.
Not yet tested
| Tool | Maker | Status |
|---|---|---|
| ChatGPT (free) | OpenAI | Not yet tested. We only rank tools we've actually run the full task set on. We never estimate a score. |
| Gemini (free) | Not yet tested. We only rank tools we've actually run the full task set on. We never estimate a score. | |
| Claude | Anthropic | Not yet tested. We only rank tools we've actually run the full task set on. We never estimate a score. |
| Microsoft Copilot | Microsoft | Not yet tested. We only rank tools we've actually run the full task set on. We never estimate a score. |
Want to see a tool tested? Tell us.