benchmarks.site

Public leaderboard · Aggregated leaderboards

Kagi LLM Benchmarking Project

Kagi, a search company, regularly quizzes AI models on a private set of tricky reasoning, coding and language questions that change over time, to see which ones actually think well.

Leaderboard

Top 3 right now

  1. 1stclaude-5-fable-thinking91.4
  2. 2ndclaude-5-fable88.8
  3. 3rdclaude-4-8-opus-thinking88.8

What it measures

Private, frequently changed set of reasoning, coding and instruction-following questions, reporting accuracy along with cost and speed.

How to read the numbers

Share of benchmark tasks answered correctly (%) — higher is better.

Results

25 rows · higher is better (Accuracy (%))CSV
1claude-5-fable-thinking91.410.4kagi (ult)
2claude-5-fable88.89.9kagi (ult)
3claude-4-8-opus-thinking88.88.8kagi (ult)
4gpt-5-588.82.2kagi (ult)
5claude-4-6-opus-thinking83.64.8kagi (ult)
6grok-4-583.51.0kagi (ult)
7claude-4-7-opus-thinking80.72.8kagi (ult)
8claude-4-5-opus-thinking80.24.5kagi (ult)
9gemini-3-pro80.10.4kagi (ult)
10kimi-k2-5-thinking78.50.5kagi
11gpt-5-pro76.831.8kagi (ult)
12glm-5-thinking75.02.0openrouter
13grok-4-20-thinking75.01.3kagi (ult)
14chatgpt-4o (points to gpt 5.3)75.00.8kagi (ult)
15claude-4-opus-thinking74.322.4kagi (ult)
16qwen-3-5-397b73.71.8together
17grok-473.61.0kagi (ult)
18gpt-5-273.31.4kagi (ult)
19claude-4-7-opus73.32.2kagi (ult)
20claude-4-sonnet-thinking73.05.4kagi (ult)
21gpt-572.77.1kagi (ult)
22qwen3-max72.51.9openrouter
23claude-4-6-opus72.41.6kagi (ult)
24o3-pro72.134.2kagi (ult)
25claude-4-5-opus70.71.4kagi (ult)

Note: Reasoning benchmark table, 116 tasks, last task list revision Feb 13, 2026; page has no explicit update date but lists models released after that. Order as published (by accuracy).

Source: Kagi Search · help.kagi.com · retrieved 2026-11-17

Markdown · More Aggregated leaderboards