SSP-Bench - Hallucination: leaderboard
The hallucination service of SSP-Bench (Deniz, Boshmaf and Khalil, QCRI, 2026), a dynamic benchmark that generates items from external sources and selects them for difficulty and separability. Each model answers 1,281 generated factual questions grounded in retrieved Wikipedia passages, and an LLM judge compares every answer with the gold answer. Non-hallucination rate in percent, counting correct answers and explicit refusals to answer; higher is better.
Metric: Non-hallucination rate (%; share of 1,281 generated Wikipedia-grounded factual questions answered correctly or explicitly declined, answers compared with the gold answer by an LLM judge; SSP-Bench dynamic benchmark instance scored on the 24-model testing panel, held out from item generation and selection; proprietary models through OpenRouter, March-April 2026). Source: arxiv.org. Saturation forecast: Estimated already saturated. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Mini | 92.9 | #349 |
| 2 | Gemini 3 Flash (Preview) | 92.7 | #141 |
| 3 | Claude Haiku 4.5 | 91.8 | #568 |
| 4 | GPT-4o | 91 | #741 |
| 5 | Qwen 3.5 122B A10B | 91 | #329 |
| 6 | Grok 3 Mini | 91 | #535 |
| 7 | Qwen 3.5 35B A3B | 90.9 | #480 |
| 8 | Claude 3.5 Haiku | 90.9 | #1563 |
| 9 | Llama 3.3 70B | 90.7 | #1511 |
| 10 | Grok 4 Fast | 90.7 | #438 |
| 11 | Qwen 3.5 27B | 90 | #358 |
| 12 | DeepSeek R1 0528 | 89.5 | #425 |
| 13 | Grok 4.1 Fast | 89.3 | #426 |
| 14 | GPT-OSS-120B | 89.1 | #699 |
| 15 | Gemma 4 31B | 89.1 | #393 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=ssp-bench-hallucination · How It Works · Data refreshed daily, snapshot 2026-11-12.