RealRun / Guides
Short answer: Most top models can schedule meetings reliably. On 20 fresh, held-back requests, Gemini 3.8 Flash, GPT-6.1 Sol, Grok 4.7, Claude Opus and Claude Sonnet all scored 100%. Grok 4.20 scored 41.7% and often booked over busy events, so we don't recommend it. GPT-5.4 mini (90%) is usable, but check every time it proposes.
Each model read a calendar and email thread, proposed a valid time and drafted a reply. Valid-answer % on 20 held-back and 42 main requests.
| Model | Held-back 20 | Main 42 | Cost |
|---|---|---|---|
| Gemini 3.8 Flash | 100.0% | 100.0% | medium |
| GPT-6.1 Sol | 100.0% | 100.0% | medium |
| Grok 4.7 | 100.0% | 100.0% | high |
| Claude Opus | 100.0% | 100.0% | subscription |
| Claude Sonnet | 100.0% | 97.6% | subscription |
| Claude Haiku | 96.7% | 97.6% | subscription |
| GPT-5.4 mini | 90.0% | 97.6% | low |
| Grok 4.20 | 41.7% | 36.5% | low |
Full scheduling results (62 requests, 1,116 graded answers).
Want the exact prompt and setup? The RealRun Meeting Scheduling Playbook (v1) has the word-for-word tested prompt, model picks by budget, and every miss explained.
Scores reflect our tests on the dates shown, not a guarantee. Claude models ran on a Claude subscription, so they have no metered API cost. cost_tier is relative within a task (low/medium/high) from our run spend; 'subscription' = not metered. Tested October 2026 by RealRun (Base Ledger IP LLC).