CTE AI Report Card
We gave free AI models the same 10 CTE teaching tasks, from an ag lesson plan to a welding safety quiz, and scored every answer 1–5 on accuracy, safety, classroom-readiness, CTE specificity and format. Every output is published in full, so you can check our grading.
Tested 2026-10-166 models · 10 tasks · 59 scored outputsHow we score →
Biggest finding: no model got the welding safety quiz fully right. All four open models keyed at least one wrong shade number or PPE answer, one of them a dangerously light Shade 5 for stick welding. Gemini was the only one with no wrong keyed answer, but it still called Shade 10 the "minimum" and left out OSHA's 20-ft cylinder separation rule. Use AI drafts as a starting point and always check safety content against OSHA, ANSI Z49.1 or your state rules before it reaches students.
The free chat apps: Gemini's free app leads overall. ChatGPT's logged-out free version scored 4.4 on 9 tasks: its welding quiz couldn't be fully captured, so it isn't scored, and that was the task every other model scored lowest on. On the 9 tasks all six models were scored on, Gemini and Qwen3.8 27B tie at 4.6 and ChatGPT scores 4.4. Both apps correctly listed all 9 major U.S. allergens, including sesame.
Rankings
| # | Model | Overall | Best CTE subject | Free tier | Privacy | |
|---|---|---|---|---|---|---|
| 1 | Gemini (free app) Google · free chat app | 4.5/5 | Cosmetology (4.8) | Free with a Google account. Google's newest model, Gemini 4 Argon, is limited to paid Google AI Pro and Ultra plans in the app. source | With Keep Activity on, Google uses chats to improve its AI and some are read by human reviewers (kept up to 3 years). Turn Keep Activity off or use a temporary chat, and never paste student names or personal info. source | Scores & full outputs → |
| 2 | ChatGPT (free, logged out) OpenAI · free chat app | 4.4/5 9 of 10 tasks scored | Cosmetology (4.8) | Free to use without an account (basic model). Free accounts get more features; paid plans unlock more powerful models. source | OpenAI may use Free-plan conversations to train its models unless you turn off 'Improve the model for everyone' (Settings > Data controls) or use a temporary chat. Never paste student names or personal info. source | Scores & full outputs → |
| 3 | Qwen3.8 27B Qwen (Alibaba) · open model via Groq | 4.4/5 | Construction (4.9) | Groq free tier: 30 requests/min, 1,000 requests/day, 8K tokens/min (our longer prompts sometimes hit the per-minute cap). 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 | GPT-OSS 120B OpenAI (open-weight) · open model via Groq | 4.2/5 | Ag & animal science (4.7) | 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 → |
| 5 | GPT-OSS 20B OpenAI (open-weight) · open model via Groq | 3.9/5 | Ag & animal science (4.5) | 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 → |
| 6 | ALLaM 2 7B SDAIA · open model via Groq | 2.0/5 | Ag & animal science (2.2) | 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). ChatGPT and Gemini were tested in their free web apps, the way a teacher would use them, so their settings are whatever the app uses. Every score is one run per task: a snapshot, not a verdict.
Not yet tested
| Tool | Maker | Status |
|---|---|---|
| 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. |
| MagicSchool | MagicSchool AI | 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. Recent AI releases and free-tier changes are on AI updates.