Longform Writing

Name: Longform Writing
Creator: EQ-Bench

EQ-Bench · Coherence and completeness of long-form stories

Official evaluation

In NewsRanked

Evidence budget2.4%

Upstream data as of09/24 15:57

Last synced10/01 20:05

What it measures, and how

Uses overall_score_100; scoring of plot and expression in an eight-chapter long text, with judge preference disclosed as a limitation.

How this evidence is used

Officially scored; the two writing leaderboards together account for 6%, with no added weight by question or model count.

Results

Under fixed rules, each public model uses one representative configuration. Anonymous test variants are not shown, original leaderboard ranks are kept, and each item lists at most 30.

Rank there	Model there	Raw score	Representative configuration
1	claude-opus-5Anthropic	86.3	Source default configuration
2	claude-fable-5-1Anthropic	85.3	Source default configuration
3	claude-opus-5-5Anthropic	84.6	Source default configuration
4	claude-fable-5Anthropic	83	Source default configuration
4	grok-4.7xAI	83	Source default configuration
6	gpt-6-solOpenAI	82.8	Source default configuration
6	gpt-6-astraOpenAI	82.8	Source default configuration
6	muse-spark-1.3Meta	82.8	Source default configuration
9	aion-3.5aion-labs	82.1	Source default configuration
10	claude-opus-4-7Anthropic	81.8	Source default configuration
10	GLM-5.3Z.ai	81.8	Source default configuration
12	gpt-5.6-solOpenAI	81.7	Source default configuration
13	muse-spark-1.2Meta	81.5	Source default configuration
14	claude-opus-4-8Anthropic	80.8	Source default configuration
15	ox-alpha—	79.9	Source default configuration
15	claude-sonnet-4-6Anthropic	79.9	Source default configuration
17	kimi-k3Moonshot AI	79.6	Source default configuration
18	moonshotai/Kimi-K2.6Moonshot AI	78.5	Source default configuration
19	claude-sonnet-5Anthropic	78.3	Source default configuration
19	gpt-5.4OpenAI	78.3	Source default configuration
21	gpt-5.5OpenAI	78.2	Source default configuration
22	gpt-5.6-terraOpenAI	78	Source default configuration
23	zai-org/GLM-5.2Z.ai	77.9	Source default configuration
24	claude-opus-4-6Anthropic	77.7	Source default configuration
25	gemini-3.8-flashGoogle	76.8	Source default configuration
26	deepseek-ai/DeepSeek-V4-ProDeepSeek	75.6	Source default configuration
27	gpt-5.6-lunaOpenAI	75.5	Source default configuration
28	moonshotai/Kimi-K2.5Moonshot AI	74.9	Source default configuration
29	Altworld/Hemmingway-1—	74.2	Source default configuration
30	deepseek-v4.1-flashDeepSeek	74	Source default configuration

Limits and data attribution

Mainly English writing, relying on a model judge; the same judge and related tasks share a budget and do not represent Chinese prose quality.

Data licence: MIT · EQ-bench-site README metadata statement

Scores published by EQ-Bench. Raw scores and the News consensus score use different scales and cannot be added directly.