Benchmark Almanac

CLadder

GPT-4 Turbo Nov 2023 70.4% leads since November 2023
Not in model cards

Models answer yes or no to questions about associations, interventions, and counterfactuals. A causal inference engine supplies the reference answers.

The makers’ repository has 149 stars. The CaDO team tested six models on a 1,989-question split that kept stories separate.

Made by
CausalNLP and collaborators
Released
May 2023
Kind
Language reasoning
Level
University
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
Subject
Mathematics
Language
English
In and out
Text → Text
Interaction
Single turn
Items from
Generated
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-4 Turbo Nov 2023 #176 on the capability index
Last model added
November 2023
Scores refreshed
November 20th, 2026
10
Entries in its paper, 7 models we track

GPT-4 Turbo leads with 70.4% among models released by November 2023 GPT-4 Turbo leads with 70.4% among models released by November 2023

Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in November 2023, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 GPT-4 Turbo Nov 2023, v1.5, CausalCoT 70.4%
  2. 2 GPT-4 Mar 2023, v1.0, CausalCoT 66.6%
  3. 3 davinci-003 v1.0, Direct prompting 62.7%
  4. 4 GPT-3.5 Turbo (Mar 2023) v1.0, Direct prompting 61.7%
  5. 5 GPT-3 Instr. (text-davinci-001) v1.0, Direct prompting 57.1%
  6. 6 davinci-002 v1.0, Direct prompting 56.2%
  7. 7 GPT-3.5 (snapshot unspecified) v1.5, Direct prompting 52.2%
  8. 8 GPT-3 175B (davinci) v1.5, Direct prompting 49.9%
  9. 9 Alpaca v1.0, Direct prompting 45.5%
  10. 10 LLaMA-7B v1.0, Direct prompting 45.2%

Source: The CLadder paper

#EntryScoreSettingDateSource
1 GPT-4 Turbo Nov 2023 70.4% v1.5, CausalCoT Nov 2023 Official leaderboard
2 GPT-4 Mar 2023 66.6% v1.0, CausalCoT Mar 2023 Official leaderboard
3 davinci-003 62.7% v1.0, Direct prompting Nov 2022 Official leaderboard
4 GPT-3.5 Turbo (Mar 2023) 61.7% v1.0, Direct prompting Mar 2023 Official leaderboard
5 GPT-3 Instr. (text-davinci-001) 57.1% v1.0, Direct prompting Official leaderboard
6 davinci-002 56.2% v1.0, Direct prompting Mar 2022 Official leaderboard
7 GPT-3.5 (snapshot unspecified) 52.2% v1.5, Direct prompting Official leaderboard
8 GPT-3 175B (davinci) 49.9% v1.5, Direct prompting May 2020 Official leaderboard
9 Alpaca 45.5% v1.0, Direct prompting Official leaderboard
10 LLaMA-7B 45.2% v1.0, Direct prompting Feb 2023 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-4 Turbo Nov 2023 70.4% v1.5, CausalCoT Nov 2023 Official leaderboard
2 GPT-4 Mar 2023 66.6% v1.0, CausalCoT Mar 2023 Official leaderboard
3 davinci-003 62.7% v1.0, Direct prompting Nov 2022 Official leaderboard
4 GPT-3.5 Turbo (Mar 2023) 61.7% v1.0, Direct prompting Mar 2023 Official leaderboard
5 davinci-002 56.2% v1.0, Direct prompting Mar 2022 Official leaderboard
6 GPT-3 175B (davinci) 49.9% v1.5, Direct prompting May 2020 Official leaderboard
7 LLaMA-7B 45.2% v1.0, Direct prompting Feb 2023 Official leaderboard

Rankings by part

We read 6 parts from the benchmark’s paper, each ranking models on one part of the test, such as Rung 1, Rung 2 and Rung 3. GPT-4 Turbo leads 4 of them, and GPT-4 leads the other 2. Arrows show how far a model moved from its overall rank among the models we track.

Rung 1

7 models we track have a score here.

  1. 1 GPT-4 Turbo Nov 2023 83.3%
  2. 2 GPT-3.5 Turbo (Mar 2023) ↑2 65.1%
  3. 3 GPT-3 175B (davinci) ↑3 63.9%
  4. 4 LLaMA-7B ↑3 63.3%
  5. 5 GPT-4 Mar 2023 ↓3 61.7%
  6. 6 davinci-003 ↓3 58%
  7. 7 davinci-002 ↓2 50.9%

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
GPT-4 Turbo
v1.5, Direc… v1.5, Causa…
70.4%
GPT-4
v1.0, Direc…
66.6%
davinci-003
v1.5, Direc… v1.0, Direc…
62.7%
davinci-002
v1.5, Direc… v1.0, Direc…
56.2%
GPT-3 175B (davinci)
v1.0, Direc…
49.9%
LLaMA-7B
v1.5, Direc…
45.2%

Source: The CLadder paper

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-4 Turbo Nov 2023 v1.5, CausalCoT 70.4%
GPT-4 Mar 2023 v1.0, CausalCoT 66.6%
GPT-4 Mar 2023 v1.0, Direct prompting 64.3%
davinci-003 v1.0, Direct prompting 62.7%
GPT-4 Turbo Nov 2023 v1.5, Direct prompting 62%
GPT-3.5 Turbo (Mar 2023) v1.0, Direct prompting 61.7%
davinci-003 v1.5, Direct prompting 56.3%
davinci-002 v1.0, Direct prompting 56.2%
davinci-002 v1.5, Direct prompting 53.1%
GPT-3 175B (davinci) v1.5, Direct prompting 49.9%
GPT-3 175B (davinci) v1.0, Direct prompting 47.4%
LLaMA-7B v1.0, Direct prompting 45.2%
LLaMA-7B v1.5, Direct prompting 44%

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
GPT-4 Turbo
v1.5, Direc… v1.5, Causa…
70.4%
GPT-4
v1.0, Direc…
66.6%
davinci-003
v1.5, Direc… v1.0, Direc…
62.7%
davinci-002
v1.5, Direc… v1.0, Direc…
56.2%
GPT-3 175B (davinci)
v1.0, Direc…
49.9%
LLaMA-7B
v1.5, Direc…
45.2%

Source: The CLadder paper

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-4 Turbo Nov 2023 v1.5, CausalCoT 70.4%
GPT-4 Mar 2023 v1.0, CausalCoT 66.6%
GPT-4 Mar 2023 v1.0, Direct prompting 64.3%
davinci-003 v1.0, Direct prompting 62.7%
GPT-4 Turbo Nov 2023 v1.5, Direct prompting 62%
GPT-3.5 Turbo (Mar 2023) v1.0, Direct prompting 61.7%
davinci-003 v1.5, Direct prompting 56.3%
davinci-002 v1.0, Direct prompting 56.2%
davinci-002 v1.5, Direct prompting 53.1%
GPT-3 175B (davinci) v1.5, Direct prompting 49.9%
GPT-3 175B (davinci) v1.0, Direct prompting 47.4%
LLaMA-7B v1.0, Direct prompting 45.2%
LLaMA-7B v1.5, Direct prompting 44%

2 models have held number one in the benchmark’s paper since its release in May 2023

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI
Benchmark paper Not tested since Nov 2023
GPT-4 Turbo 70.4%

Source: The CLadder paper

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2020 2021 2022 2023 2024 Near the ceiling (85%) GPT-4 Turbo GPT-4 davinci-002 GPT-3 175B (davinci) 0% 20% 40% 60% 80% 100% 2020 2021 2022 2023 2024 Near the ceiling (85%) GPT-4 Turbo davinci-003

Source: Official CLadder leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Knowledge

All 332 in Knowledge
  • MMLU-Pro In use Harder 10-option version of MMLU with reasoning-focused questions across 14 subjects. Intern-S2-397B 89.8%
  • MMLU In use 57-subject multiple-choice exam questions from elementary to professional level. Scored by accuracy. Seed1.8 92.3%
  • SuperGPQA In use 26,529 graduate-level multiple-choice questions across 285 disciplines. Accuracy. Qwen 3.6 Max (Preview) 73.9%
  • WideSearch In use 200 bilingual broad info-seeking tasks: agent collects many atomic facts into a table. Scored by success rate and row/item F1. Tencent Hy4 preview 83.9%
  • BBEH micro average In use Models solve harder replacements for each of the 23 BIG-Bench Hard tasks, 4,520 examples in total. The score is the share of all examples answered correctly. Gemma 4 31B IT 74.4%
  • HLE-Diamond In use Models answer 1,000 cleaned expert questions from Humanity's Last Exam, 500 on reasoning and 500 on knowledge, closed book. Graded on exact answers. Claude Fable 5.2 77.9%

Data audit

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 3 years ago, in November 2023, for GPT-4 Turbo. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the benchmark’s paper (10 entries).
  • Pass Models scored 7 models we track, plus 3 other entries on its leaderboard.
  • Warning Newest score The newest came 3 years ago, in November 2023, for GPT-4 Turbo.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-4 Turbo (#176).
  • Pass Headroom The best is 70.4%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 13 rows for models we track and 3 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check