How we test and score
1. The same 10 tasks for every model
Each model gets the exact prompts below, with the same system message ("You are helping a school-based speech-language pathologist (SLP). Produce ready-to-use material."), temperature 0.3 and a 3,000-token output budget (reasoning models spend part of that budget thinking). If a first run came back empty, we re-ran it once with a 6,000-token budget, and that output page says so. Outputs that hit the length limit are scored as delivered, cut off. No task contains any student's personal information. Exception: Groq's free tier limits Qwen3.8 27B to 1,000 output tokens per minute and rejects requests it expects to exceed that, so some Qwen tasks ran with a 1,500-token budget. Each output page shows the budget used.
1. Articulation /r/ activity Articulation
Create a 30-minute small-group articulation activity for 3 second-grade students working on vocalic /r/ ("er" as in bird, her, teacher) at the word and phrase level. Include a target word list of 20 words sorted by position (initial, medial, final), a game that needs only paper and a die, data-collection instructions (trials and accuracy), and one home-practice idea for families. Use only real words that actually contain the target sound.2. /s/-blend word list Articulation
Write a list of 30 real, picturable words for a kindergarten-1st grade student working on initial /s/-blends: 10 with /st/, 10 with /sp/, 10 with /sn/ or /sm/ (5 each). Then write 6 short carrier sentences that use those words. Double-check that every word starts with the stated blend.
3. Expressive language IEP goals Language
Write 4 measurable annual IEP goals for a 4th-grade student with an expressive language disorder (short, simple sentences; limited use of conjunctions; difficulty retelling stories). Each goal should state the condition, the observable behavior, the criterion (accuracy and number of sessions or probes), and how it will be measured. Leave blanks like [Student] and [date] instead of names. Add one short-term objective under each goal.
4. Progress-note template IEP & documentation
Create a fill-in-the-blank quarterly IEP progress report template a school SLP can reuse for any student on their caseload. Include blanks for goal text, baseline, current performance data, progress code (with a short legend such as Mastered / Sufficient progress / Insufficient progress / Not yet introduced), and narrative comments, plus 3 example narrative sentences written with placeholders like [Student]. Do not include any real or made-up student names.
5. Social pragmatics lesson Social pragmatics
Plan a 30-minute social communication group session for three 6th graders working on conversational turn-taking and staying on topic. Include a warm-up, a structured activity with visual supports described in words, 3 role-play scenarios appropriate for middle school, how the SLP will prompt and fade cues, and a simple data sheet layout. Use a neurodiversity-affirming tone (do not frame eye contact as required).
6. AAC core-word plan AAC
Write a one-week plan for modeling AAC core vocabulary with a 2nd-grade student who uses a robust symbol-based speech-generating device. Pick 5 core words for the week, explain what aided language modeling is, give 2 classroom routines per day where each word can be modeled, and include a short handout for the classroom teacher and paraprofessional (no device brand required). Do not require the student to show prerequisite skills before getting access.
7. Fluency session plan Fluency
Write a 30-minute therapy session plan for a 5th-grade student who stutters. Include goals that address both fluency-shaping or stuttering-modification techniques and communication attitudes, a description of one technique (such as easy onset or pull-outs) explained in student-friendly language, an activity that practices it, and a short guide for the classroom teacher on how to respond when the student stutters. Avoid telling the student to 'slow down' or 'take a breath' as advice from listeners.
8. Phonological awareness quiz Language
Create a 10-item phonological awareness screening probe for a first grader that an SLP can read aloud: 3 rhyme items, 3 initial-sound isolation items, 2 syllable-counting items and 2 phoneme-blending items. Provide an answer key with the correct response for each item and a simple scoring line. Make sure every keyed answer is correct.
9. Sound development parent handout Articulation
Write a one-page parent handout explaining typical ages of speech sound development for English (for example /p, b, m/ early, /s, r, th/ later) and when to ask the school SLP about concerns. Cite which research or organization the age ranges come from (for example, ASHA or a named study) and note that norms vary by source. Keep it at about a 6th-grade reading level.
10. Evaluation summary template IEP & documentation
Create a template for the summary and recommendations section of a school speech-language evaluation report. Include placeholder blanks for standardized test names, standard scores, percentile ranks, and confidence intervals, a sentence explaining that standard scores have a mean of 100 and standard deviation of 15, a short guide to interpreting scores (e.g., which ranges are average or below average), an eligibility statement with blanks, and recommendations. Use only placeholders like [Student] and [Test Name], no real names or invented scores.
2. The rubric (1–5 on each criterion)
| Accuracy | Facts, word lists, norms, score math and answer keys are correct (e.g., every target word actually contains the target sound). |
| Clinical soundness | Reflects current evidence-based, student-centered practice: measurable goals, no prerequisite gatekeeping for AAC, neurodiversity-affirming pragmatics, no 'slow down' advice for stuttering. No student personal info. |
| Therapy-readiness | An SLP could use it tomorrow with light edits: follows the requested count, length and structure. |
| SLP specificity | Uses real school-SLP practice (cueing hierarchies, trials and accuracy data, IEP/progress-report language, ASHA-style terms) instead of generic filler. |
| Format | Clean, printable layout; complete (not cut off); blanks and answer keys where asked. |
5 = no errors found, ready to use. 4 = minor slip. 3 = one real error or a notable gap to fix. 2 = several errors, or one that could mislead students. 1 = unusable. A task score is the average of the five criteria, and the overall score is the average of the task scores. Best area is the subject area whose tasks scored highest.
3. Disclosure
Testing and scoring are done by SLPPrompts's Grok bots, the AI assistants that run this site. Each output was read in full against the rubric, answer keys were checked item by item, and every deduction is explained on the model's page next to the full, unedited output, so you can disagree with us. We don't accept payment for scores. Free-tier and privacy notes link to the provider's own page. Scoring is a judgment call, not a benchmark. If you spot a grading mistake, tell us and we'll fix it and note the change.
4. What we didn't test yet
The free chat apps (ChatGPT, Gemini, Claude and Microsoft Copilot) are listed as "Not yet tested" until we run the full task set on them. We never estimate a score.