BenchLeader

EQ-Bench Longform Writing leaderboard

As of 21 Nov 2026, the top score on EQ-Bench Longform Writing is 86.3 by Claude Opus 5 (Anthropic), ahead of Claude Fable 5.1 at 85.3, across 125 model configurations with a published result (EQ-Bench). The model plans a short story or novella from a minimal prompt, revises the plan, then writes it in eight chapters of about 1,000 words; an LLM judge scores each chapter and the finished piece against a rubric, averaged on a 0–100 scale. EQ-Bench Longform Writing is shown for reference only and does not count toward the BenchLeader Index.

Key facts (as of 21 Nov 2026)

  • Claude Opus 5 holds the top EQ-Bench Longform Writing score at 86.3 (EQ-Bench, 21 Nov 2026).
  • Claude Fable 5.1 is second on EQ-Bench Longform Writing at 85.3 (EQ-Bench, 21 Nov 2026).
  • EQ-Bench Longform Writing: The model plans a short story or novella from a minimal prompt, revises the plan, then writes it in eight chapters of about 1,000 words; an LLM judge scores each chapter and the finished piece against a rubric, averaged on a 0–100 scale (EQ-Bench).
  • 125 model configurations have a published EQ-Bench Longform Writing result on BenchLeader (BenchLeader, 21 Nov 2026).
  • EQ-Bench Longform Writing does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).

The model plans a short story or novella from a minimal prompt, revises the plan, then writes it in eight chapters of about 1,000 words; an LLM judge scores each chapter and the finished piece against a rubric, averaged on a 0–100 scale.

Published by
EQ-Bench
Category
Human preference
Index weight
Reference only
Models
125
Data as of
21 Nov 2026

MIT — EQ-Bench.

125 of 125
#
1Claude Opus 586.363.2
2Claude Fable 5.185.366.9
3Claude Opus 5.584.662.6
4Claude Fable 58365.1
5Grok 4.783–
6GPT-6 Astra82.868.0
7Muse Spark 1.382.8–
8GPT-6 Sol82.8–
9Claude Opus 4.781.863.0
10GLM 5.381.8–
11GPT-5.6 Sol81.756.4
12Muse Spark 1.281.5–
13Gemini 4 Argon80.963.3
14Claude Opus 4.880.858.5
15Kimi K3.1 Pro79.960.4
16Claude Sonnet 4.679.958.3
17Ox Alpha79.9–
18Kimi K379.660.6
19Kimi K2.678.555.0
20GPT-5.478.357.7
21Claude Sonnet 578.352.5
22GPT-5.578.260.8
23GPT-5.6 Terra78–
24GLM 5.277.951.8
25Claude Opus 4.677.760.7
26Gemini 3.8 Flash76.855.7
27DeepSeek V4 Pro75.652.8
28GPT-5.6 Luna75.547.0
29Kimi K2.574.954.2
30DeepSeek V4.1 Flash74–
31MiMo-V2.5-Pro73.755.5
32GLM 5.173.555.2
33Claude Opus 4.573.156.9
34GPT-5.4 Mini73–
35GPT-572.856.8
36Inkling72.550.2
37Gemini 3.5 Flash71.8–
38GPT-5.271.655.4
39Claude Sonnet 4.571.453.6
40Mimo v2 Pro70.959.0
41GLM-570.954.8
42Muse Spark 1.170.661.6
43Claude Opus 4.170.551.7
44Kimi K2 Thinking6952.2
45Gemini 3 Pro68.860.3
46Grok 4.2068.5–
47Qwen3.5 397B A17B68.351.2
48Gemini 3.1 Pro68.259.5
49Kimi K268.151.0
50DeepSeek V3.266.848.7
51Qwen3 Max66.253.2
52DeepSeek V4 Flash6653.7
53GLM 4.76651.6
54MiniMax M2.565.353.6
55Claude Haiku 4.56547.6
56Kimi K2 090563.851.7
57o362.760.8
58GPT-5.3 Chat62.658.2
59DeepSeek V3.162.348.5
60ChatGPT-4o60.254.7

More human preference benchmarks · All LLM benchmarks

Cite as: BenchLeader, “EQ-Bench Longform Writing leaderboard”, data as of 21 Nov 2026, https://www.benchleader.com/benchmarks/eqbench_longform_writing.

EQ-Bench Longform Writing: questions

What does EQ-Bench Longform Writing measure?
The model plans a short story or novella from a minimal prompt, revises the plan, then writes it in eight chapters of about 1,000 words; an LLM judge scores each chapter and the finished piece against a rubric, averaged on a 0–100 scale. Scores are reported in score; higher is better.
What is the top score on EQ-Bench Longform Writing?
As of 21 Nov 2026, the highest score on EQ-Bench Longform Writing is 86.3 by Claude Opus 5, ahead of Claude Fable 5.1 at 85.3. Scores come from EQ-Bench and are re-read daily.
Which AI model leads EQ-Bench Longform Writing?
Claude Opus 5 leads EQ-Bench Longform Writing with 86.3 as of 21 Nov 2026, ahead of Claude Fable 5.1 at 85.3.
How many models have EQ-Bench Longform Writing results?
125 model configurations have a EQ-Bench Longform Writing result on BenchLeader, all taken from EQ-Bench.
Who runs EQ-Bench Longform Writing and how often is it updated?
EQ-Bench Longform Writing is published by EQ-Bench. BenchLeader re-reads the published results once a day and records the date each result was published.
Does EQ-Bench Longform Writing count toward the BenchLeader Index?
No. EQ-Bench Longform Writing is shown for reference but left out of the composite index.