Skip to content

Benchmark

FaithJudge Hallucination Rate RagTruth (Summarization)

Factuality benchmarks measure how often models get facts right. Scores are percentages. Lower is better.

Models and systems measured
43
Scores recorded
43
Sources
1
IQ weight
1.87

Score distribution

18.8%
Mean
12.7%
Median
15.3%
Std dev
8%
10th percentile
30%
90th percentile
4.7%
Min
4.7%74%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.87 for chat models, measured over 31 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model or systemScore
1Gemini 2.5 Flashgemini-2.5-flash, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data4.7%
2Gemini 2.5 Progemini-2.5-pro, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data4.7%
3Gemini 2.0 Flashgemini-2.0-flash-001, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data6.7%
4Phi 4phi-4, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data8%
5GPT-3.5 Turbogpt-3.5-turbo-0125, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data8.7%
6Llama 3.3 70B InstructLlama-3.3-70B-Instruct, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data8.7%
7Gemini 2.5 Progemini-2.5-pro-exp-03-25, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data9.3%
8GPT 4ogpt-4o-2024-11-20, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data10%
9GPT-4.5 Previewgpt-4.5-preview-2025-02-27, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data10%
10Mistral Small 24B InstructMistral-Small-24B-Instruct-2501, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization), 2501· FaithJudge data10%
11GPT-4.1 minigpt-4.1-mini-2025-04-14, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data10.7%
12GPT 4.1gpt-4.1-2025-04-14, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data11.3%
13GPT-4o minigpt-4o-mini-2024-07-18, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data11.3%
14Grok 3grok-3, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data12%
15Claude Opus 4claude-opus-4-20250514, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data12.7%
16Llama 3.1 8B InstructLlama-3.1-8B-Instruct, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data12.7%
17Qwen2.5 72B InstructQwen2.5-72B-Instruct, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data12.7%
18Claude Sonnet 4claude-sonnet-4-20250514, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data13.3%
19Llama 4 Maverickllama-4-maverick, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data13.3%
20Qwen2.5 32B InstructQwen2.5-32B-Instruct, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data13.3%
21Claude Sonnet 3.7claude-3-7-sonnet-20250219, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data14.7%
22Qwen 2.5 14B InstructQwen2.5-14B-Instruct, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data18%
23Phi 4 MiniPhi-4-mini-instruct, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data21.3%
24jamba-1.6-miniAI21-Jamba-mini-1.6, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data25.3%
25Qwen2.5 7B InstructQwen2.5-7B-Instruct, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data26%
26Llama 3.2 3BLlama-3.2-3B-Instruct, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data28.7%
27QwQ 32BQwQ-32B, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data30%
28Qwen2.5 3B InstructQwen2.5-3B-Instruct, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data42%
29Qwen2.5 1.5B InstructQwen2.5-1.5B-Instruct, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data60.7%
30Llama 3.2 1B InstructLlama-3.2-1B-Instruct, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data64.7%
31Qwen2.5 0.5b InstructQwen2.5-0.5B-Instruct, o3-mini-high, FaithJudge RAG generation hallucination rate; human-example-guided LLM judge, RagTruth (Summarization)· FaithJudge data74%

Not ranked

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
claude-3-7-sonnet-thinking-20250219thinking, judge o3-mini-highthinking· FaithJudge data10.7%
claude-opus-4-thinking-20250514thinking, judge o3-mini-highthinking· FaithJudge data11.3%
claude-sonnet-4-thinking-20250514thinking, judge o3-mini-highthinking· FaithJudge data15.3%
o3-high-2025-04-16reasoning effort high, judge o3-mini-highreasoning effort high· FaithJudge data20.7%
o3-low-2025-04-16reasoning effort low, judge o3-mini-highreasoning effort low· FaithJudge data22.7%
o3-medium-2025-04-16reasoning effort medium, judge o3-mini-highreasoning effort medium· FaithJudge data19.3%
o3-mini-high-2025-01-31reasoning effort high, judge o3-mini-highreasoning effort high· FaithJudge data8%
o3-mini-low-2025-01-31reasoning effort low, judge o3-mini-highreasoning effort low· FaithJudge data6.7%
o3-mini-medium-2025-01-31reasoning effort medium, judge o3-mini-highreasoning effort medium· FaithJudge data9.3%
o4-mini-high-2025-04-16reasoning effort high, judge o3-mini-highreasoning effort high· FaithJudge data23.3%
o4-mini-low-2025-04-16reasoning effort low, judge o3-mini-highreasoning effort low· FaithJudge data30%
o4-mini-medium-2025-04-16reasoning effort medium, judge o3-mini-highreasoning effort medium· FaithJudge data23.3%

Data credits: FaithJudge, Vectara; Tamber et al., EMNLP 2025 Industry. Public numeric README leaderboard; results licence not separately established.