SpeechMap Compliance: leaderboard
Models are sent sensitive and controversial prompts spanning political argument, satire, religion, history and rights advocacy, and a judge model labels each reply complete, evasive, a denial or a provider error. The board reports the share answered in full, so it tracks willingness to engage rather than capability, and higher means fewer refusals and hedges.
Metric: % Requests Completed. Source: speechmap.ai. 411 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok 4 | 98.3 | #181 |
| 2 | Mistral Large 3 | 98.2 | #401 |
| 3 | Grok 4.20 (Reasoning) | 98.2 | #175 (Grok 4.20) |
| 4 | Grok 4.1 Fast (Non-reasoning) | 97.9 | #217 (Grok 4.1 Fast) |
| 5 | Grok 3 | 96.1 | #304 |
| 6 | Grok 2 (1212) | 95.9 | #496 |
| 7 | Grok 4.1 Fast (Reasoning) | 95.7 | #217 (Grok 4.1 Fast) |
| 8 | GPT-4 (0314) | 95.2 | #574 |
| 9 | GPT-3.5 Turbo (0125) | 94.5 | #984 |
| 10 | Mistral Small 3.2 | 93.7 | #605 |
| 11 | Grok Beta | 93.7 | #487 |
| 12 | Ministral 3 14B | 93.3 | #650 |
| 13 | Grok 4.5 | 93.1 | #48 |
| 14 | Magistral Medium | 92.6 | #494 |
| 15 | Grok 3 Mini Beta | 92.5 | #296 |
Interactive version: theaggregate.ai/benchmark?slug=speechmap-compliance · How It Works · Data refreshed daily, snapshot 2026-10-20.