benchmarks.site

Public leaderboard · General knowledge & reasoning

SimpleBench

Trick questions about everyday situations, space, time and people that ordinary adults usually get right but that can trip up AI. It tests common sense rather than book knowledge.

Leaderboard

Top 3 right now

  1. 1stClaude Fable 5.2 (unknown)94.9
  2. 2ndGemini 4 Carbon (unknown)92.6
  3. 3rdGemini 4 Argon (unknown)91.7

What it measures

Multiple-choice trick questions on spatial and temporal reasoning, social situations and common sense that non-expert humans answer well, scored by accuracy against a human baseline.

How to read the numbers

Share of questions answered correctly, averaged over 5 runs (%); the human baseline is about 84% — higher is better.

Results

25 rows · higher is better (Score AVG@5 (%))CSV
1Claude Fable 5.2 (unknown)Anthropic94.92026-11-09
2Gemini 4 Carbon (unknown)Google DeepMind92.62026-11-10
3Gemini 4 Argon (unknown)Google DeepMind91.72026-10-13
4Claude Opus 5.5 (unknown)Anthropic88.42026-09-22
5Claude Fable 5.1 (unknown)Anthropic86.62026-09-01
6GPT-6 Astra (pro, max)OpenAI86.52026-09-03
7GPT-6 Astra (unknown thinking)OpenAI83.62026-09-03
8GPT-6.1 Sol (unknown thinking)OpenAI82.92026-09-29
9Gemini 3.8 Flash (unknown)Google DeepMind82.42026-09-02
10Claude Fable 5 (max)Anthropic81.92026-06-09
11Muse Spark 1.3 (unknown)Meta AI81.82026-09-02
12Claude Opus 5 (unknown)Anthropic80.62026-07-24
13Gemini 3.1 Pro PreviewGoogle DeepMind79.62026-02-19
14GPT-5.5 Pro (unknown thinking)OpenAI76.92026-04-23
15GPT-5.5 Pro (high)OpenAI76.92026-04-23
16Gemini 3.5 FlashGoogle DeepMind76.72026-05-19
17Gemini 3 Pro PreviewGoogle DeepMind76.42025-11-18
18Grok 4.6 (unknown)xAI75.92026-08-12
19Claude Sonnet 5.5 (unknown)Anthropic75.92026-09-28
20Muse Spark 1.2Meta AI74.52026-08-05
21GPT-5.4 ProOpenAI74.12026-03-05
22GPT-6 Sol (unknown thinking)OpenAI73.12026-09-22
23Muse Ember (unknown)Meta AI72.72026-11-09
24GPT-6.1 Terra (unknown thinking)OpenAI72.22026-10-26
25GPT-5.6 Sol (pro, xhigh)OpenAI71.72026-07-09

Note: Scores published on simple-bench.com, as mirrored by Epoch AI (simplebench_external.csv), converted to % (1 decimal). '(unknown)' means the reasoning-effort setting was not stated. Human baseline figure is from simple-bench.com.

Source: Epoch AI (Benchmarking Hub) · epoch.ai · results as of 2026-11-17 · retrieved 2026-11-17 · license: CC-BY-4.0

Markdown · More General knowledge & reasoning