LiveFact: leaderboard

LiveFact (2026) is a monthly fake-news verification benchmark built from fresh news events: models label generated claims Real, Fake or Ambiguous given evidence available three days before, on, or three days after each event. This board is the paper average of twelve scores (accuracy and macro-F1 in two modes at three evidence offsets) on the November 2025 release, in percent; higher is better.

Metric: Average (%) of the twelve LiveFact November 2025 scores (accuracy and macro-F1 in Classification and Inference modes at evidence offsets of -3, 0 and +3 days), 4,392 claims from 737 news events, temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 3 235B A22B 2507 Instruct72.4#616
2GPT-OSS-120B72.13#704
3GPT-5.1 (2025-11-13)72.02#191
4GPT-5.271.52#185
5Qwen 3 30B A3B 2507 Instruct69.46#1100
6GPT-4o (2024-08-06)67.11#743
7GPT-4o Mini (2024-07-18)66.12#1484
8Qwen 3 32B66.03#972
9Qwen 3 8B63.62#1873
10DeepSeek V3.161.48#511
11GPT-OSS-20B57.05#1183
12Llama 3.3 70B Instruct55.85#1385
13Qwen 3 4B 2507 Instruct55.46#2221
14Kimi K2 (Thinking)54.6
15Llama 3.1 8B Instruct49.96#4600

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=livefact · How It Works · Data refreshed daily, snapshot 2026-11-21.