BenchLeader

DTBench leaderboard

As of 15 Nov 2026, the top score on DTBench is 99.2% by Claude Fable 5.2 (Anthropic), ahead of Claude Opus 5.5 at 98.9%, across 238 model configurations with a published result (Epoch AI Benchmarking Hub, result dated 9 Nov 2026). Decision problems where the model's answer preference is compared with decision-theoretic expert answers. DTBench is shown for reference only and does not count toward the BenchLeader Index.

Key facts (as of 15 Nov 2026)

  • Claude Fable 5.2 holds the top DTBench score at 99.2% (Epoch AI Benchmarking Hub, 9 Nov 2026).
  • Claude Opus 5.5 is second on DTBench at 98.9% (Epoch AI Benchmarking Hub, 22 Sep 2026).
  • DTBench: Decision-theory dilemmas scored against expert answers (Epoch AI Benchmarking Hub).
  • 238 model configurations have a published DTBench result on BenchLeader (BenchLeader, 15 Nov 2026).
  • DTBench does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).

Decision-theory dilemmas scored against expert answers.

Published by
Epoch AI Benchmarking Hub
Category
Reasoning
Index weight
Reference only
Models
238
Data as of
15 Nov 2026
Newest result
9 Nov 2026

CC BY 4.0 — Epoch AI, ‘AI Benchmarking Hub’, epoch.ai/benchmarks. Mirrored boards credit their original publishers.

What the test looks like

Decision problems where the model's answer preference is compared with decision-theoretic expert answers.

How it is scored

Accuracy, as published by DTBench.

What to keep in mind

Philosophy-of-decision framing; a large model set.

238 of 238
#
1NewClaude Fable 5.299.2%65.42026-11-09
2NewClaude Fable 5.2xhigh98.9%69.72026-11-09
3Claude Opus 5.5max98.9%68.42026-09-22
4Claude Fable 5max98.4%65.12026-06-09
5NewClaude Fable 5.2high98.1%68.42026-11-09
6NewClaude Fable 5.2medium98.1%66.32026-11-09
7NewClaude Fable 5.2low98.1%65.32026-11-09
8Claude Opus 5high97.9%64.42026-07-24
9Claude Opus 5xhigh97.9%60.12026-07-24
10Claude Fable 5.197.6%66.42026-09-01
11Claude Opus 5max97.6%65.02026-07-24
12Claude Fable 5.1xhigh97.6%64.62026-09-01
13Gemini 4 Argonlow97.6%–2026-10-13
14GPT-6 Astraxhigh97.3%66.02026-09-03
15Claude Opus 5.5xhigh97.3%65.42026-09-22
16Gemini 4 Argonhigh97.3%65.32026-10-13
17Claude Sonnet 5.5max97.3%64.22026-09-28
18GPT-6 Solmax97.3%62.62026-09-22
19Gemini 4 Argonmedium97.3%61.62026-10-13
20Grok 4.6xhigh97.3%56.12026-08-12
21Claude Fable 5xhigh97.3%–2026-06-09
22GPT-6 Astra97.1%66.42026-09-03
23GPT-6 Astramedium97.1%64.42026-09-03
24Claude Fable 5.1medium97.1%61.72026-09-01
25Gemini 3.1 Promedium97.1%–2026-02-19
26Gemini 3.7 Flashhigh96.8%62.12026-08-13
27GPT-6.1 Solmax96.5%66.62026-09-29
28GPT-6 Astrahigh96.5%66.32026-09-03
29Claude Opus 5.5high96.5%66.02026-09-22
30Claude Fable 5high96.5%63.92026-06-09
31Claude Fable 5.1high96.5%63.72026-09-01
32Muse Spark 1.396.5%61.62026-09-02
33Claude Opus 5.5medium96.5%61.22026-09-22
34Grok 4.5high96.5%56.82026-07-08
35GPT-5.5xhigh96.0%64.12026-04-23
36Claude Sonnet 5.5xhigh96.0%62.52026-09-28
37Grok 4.7xhigh96.0%57.42026-09-21
38GPT-5.5 Proxhigh96.0%–2026-04-23
39GPT-5.6 Solpro mode96.0%–2026-07-09
40GPT-5.6 Solxhigh95.7%63.92026-07-09
41Claude Fable 5medium95.7%–2026-06-09
42Gemini 3.8 Flash95.7%56.42026-09-02
43GPT-5.6 Solmax95.5%65.32026-07-09
44GPT-5.6 Solhigh95.5%63.42026-07-09
45Gemini 3.6 Flashhigh95.5%61.02026-07-21
46Gemini 3.1 Prohigh95.5%60.42026-02-19
47Claude Opus 5.5low95.5%58.92026-09-22
48Muse Wonder95.2%64.42026-10-12
49GPT-5.6 Solmedium95.2%61.32026-07-09
50Claude Haiku 5.5max95.0%55.32026-10-07
51GPT-6 Astralow94.9%62.02026-09-03
52Claude Opus 4.8max94.9%61.12026-05-28
53Gemini 3.8 Flashmedium94.9%60.22026-09-02
54Claude Opus 5medium94.9%57.52026-07-24
55Gemini 3.8 Flashlow94.9%55.92026-09-02
56Gemini 3.1 Prolow94.9%–2026-02-19
57Gemini 3.5 Flashhigh94.7%62.52026-05-19
58Muse Spark 1.2xhigh94.7%61.52026-08-05
59Gemini 3.7 Flashmedium94.7%60.12026-08-13
60Claude Opus 4.7max94.7%59.72026-04-16

AI reasoning and math benchmark leaderboard · All LLM benchmarks

Cite as: BenchLeader, “DTBench leaderboard”, data as of 15 Nov 2026, https://www.benchleader.com/benchmarks/dtbench.

DTBench: questions

What does DTBench measure?
Decision problems where the model's answer preference is compared with decision-theoretic expert answers. Scores are reported in percent of tasks solved; higher is better.
What is the top score on DTBench?
As of 15 Nov 2026, the highest score on DTBench is 99.2% by Claude Fable 5.2 (result dated 9 Nov 2026), ahead of Claude Opus 5.5 at 98.9%. Scores come from Epoch AI Benchmarking Hub and are re-read daily.
Which AI model leads DTBench?
Claude Fable 5.2 leads DTBench with 99.2% as of 15 Nov 2026 (result dated 9 Nov 2026), ahead of Claude Opus 5.5 at 98.9%.
How many models have DTBench results?
238 model configurations have a DTBench result on BenchLeader, all taken from Epoch AI Benchmarking Hub.
Who runs DTBench and how often is it updated?
DTBench is published by Epoch AI Benchmarking Hub. BenchLeader re-reads the published results once a day and records the date each result was published.
Does DTBench count toward the BenchLeader Index?
No. DTBench is shown for reference but left out of the composite index.