LiveFact - Inference Macro-F1: 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 macro-F1 in percent when the label is what the evidence from three days after the event supports (Inference Mode, November 2025 release); higher is better.

Metric: Macro-F1 (%) in Inference Mode (label adjusted to what the evidence slice supports) with the evidence available three days after the event, LiveFact November 2025; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5.1 (2025-11-13)81.99#191
2Qwen 3 235B A22B 2507 Instruct81.94#616
3GPT-OSS-120B81.31#704
4GPT-5.279.57#185
5Qwen 3 30B A3B 2507 Instruct76.4#1100
6GPT-4o (2024-08-06)72.98#743
7Qwen 3 32B69.91#972
8GPT-4o Mini (2024-07-18)69.45#1484
9Qwen 3 8B68.3#1873
10DeepSeek V3.166.38#511
11Qwen 3 4B 2507 Instruct64.27#2221
12Llama 3.3 70B Instruct63.93#1385
13GPT-OSS-20B62.34#1183
14Kimi K2 (Thinking)59.15
15Llama 3.1 8B Instruct45.14#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-inference-macro-f1 · How It Works · Data refreshed daily, snapshot 2026-11-21.