AI Accuracy Scoreboard
Live rankings based on real multi-model consensus data.
| # | Model | Accuracy |
|---|---|---|
| 1 | Gemini 3.5 | 72% |
| 2 | Claude Opus 4.8 | 72% |
| 3 | Gemini 2.5 Flash | 71% |
| 4 | GPT 5.5 | 64% |
| 5 | Grok 3 | 64% |
| 6 | Grok 4 | 62% |
Insufficient data
These models have fewer than 100 verified claims so far and are excluded from the ranking. They will appear above once enough data is collected.
- Claude Haiku 4.5 66 claims
- Gemini 2.5 Flash Lite 63 claims
- Claude Sonnet 4.6 59 claims
- GPT-4o mini 59 claims
- Grok 3 Mini 52 claims
- GPT-4o 51 claims
Accuracy by Category
Categories with fewer than 50 verified claims are hidden due to insufficient data.
| # | Model | Accuracy | Claims |
|---|---|---|---|
| 1 | Gemini 2.5 Flash | 100% | 31 |
| 2 | GPT-4o | 100% | 31 |
| 3 | Claude Sonnet 4.6 | 95% | 38 |
| 4 | Grok 3 | 90% | 39 |
| 5 | GPT-4o mini | 79% | 14 |
| 6 | Grok 3 Mini | 75% | 12 |
| 7 | Claude Haiku 4.5 | 67% | 21 |
| 8 | Gemini 2.5 Flash Lite | 54% | 13 |
Methodology
Accuracy is measured by cross-model consensus. A model is accurate when its claims are corroborated by other independent models. Each question is sent to multiple AI models simultaneously, and their answers are compared at the claim level using algorithmic semantic matching.
Accuracy varies by question type and model version. Rankings reflect data collected through NoParrot.
Contribute to the scoreboard
Every question you ask helps build more accurate rankings. Try NoParrot and see how AI models compare on your questions.
Ask a question