Writing & Preference benchmark

EQ-Bench Creative Writing leaderboard

As of October 2026, GPT-6 Astra has the highest published EQ-Bench Creative Writing score on Noometry at 2173, out of 115 models with results.

Last verified October 10, 2026

About EQ-Bench Creative Writing

Elo ratings from pairwise comparisons of short stories written to creative prompts, judged by an LLM against a detailed rubric (v3).

Category: Writing & Preference
Introduced: 2025
Format: Pairwise LLM-judged stories
Unit: Arena rating (Bradley–Terry)
Official site: eqbench.com

Top 15 models

GPT-6 Astra 2173
Claude Fable 5.1 2162
Claude Opus 5 2133
GPT-6 Sol 2125
Kimi K3 2082
GLM-5.3 2075
Claude Opus 5.5 2050
Grok 4.7 2007
GPT-5.6 Sol 1972
Claude Fable 5 1943
Aion 3.5 1929
Muse Spark 1.1 1927
Claude Opus 4.7 1914
Muse Spark 1.3 1906
GPT-5.6 Terra 1855
170018001900200021002200

Sponsored placements are available on pages like this one. Advertise on Noometry

All results

EQ-Bench Creative Writing results by model
#	Model	Provider	Rating	Source
1	GPT-6 Astra	OpenAI	2173	EQ-Bench
2	Claude Fable 5.1	Anthropic	2162	EQ-Bench
3	Claude Opus 5	Anthropic	2133	EQ-Bench
4	GPT-6 Sol	OpenAI	2125	EQ-Bench
5	Kimi K3	Moonshot AI	2082	EQ-Bench
6	GLM-5.3	Z.ai (Zhipu)	2075	EQ-Bench
7	Claude Opus 5.5	Anthropic	2050	EQ-Bench
8	Grok 4.7	xAI	2007	EQ-Bench
9	GPT-5.6 Sol	OpenAI	1972	EQ-Bench
10	Claude Fable 5	Anthropic	1943	EQ-Bench
11	Aion 3.5	AionLabs	1929	EQ-Bench
12	Muse Spark 1.1	Meta	1927	EQ-Bench
13	Claude Opus 4.7	Anthropic	1914	EQ-Bench
14	Muse Spark 1.3	Meta	1906	EQ-Bench
15	GPT-5.6 Terra	OpenAI	1855	EQ-Bench
16	GPT-5.5	OpenAI	1844	EQ-Bench
17	Qwen3.8 2.4T A95B	Alibaba (Qwen)	1843	EQ-Bench
18	Muse Spark 1.2	Meta	1840	EQ-Bench
19	Claude Opus 4.8	Anthropic	1840	EQ-Bench
20	GPT-5.4	OpenAI	1840	EQ-Bench
21	GPT-5.6 Luna	OpenAI	1829	EQ-Bench
22	Claude Sonnet 4.6	Anthropic	1810	EQ-Bench
23	Claude Opus 4.6	Anthropic	1809	EQ-Bench
24	Muse Glimmer 30B	Meta	1798	EQ-Bench
25	Claude Sonnet 5	Anthropic	1794	EQ-Bench
26	GLM-5.2	Z.ai (Zhipu)	1757	EQ-Bench
27	Gemini 3.8 Flash	Google	1748	EQ-Bench
28	Kimi K2.6	Moonshot AI	1725	EQ-Bench
29	Gemini 3.7 Flash	Google	1723	EQ-Bench
30	GPT-5.2	OpenAI	1703	EQ-Bench
31	Nemotron 3 Ultra	NVIDIA	1692	EQ-Bench
32	GPT-5.3 Chat	OpenAI	1690	EQ-Bench
33	Claude Opus 4.5	Anthropic	1687	EQ-Bench
34	Claude Sonnet 4.5	Anthropic	1678	EQ-Bench
35	o3	OpenAI	1676	EQ-Bench
36	Qwen3.8 27B	Alibaba (Qwen)	1671	EQ-Bench
37	Kimi K2 (Jul 2025)	Moonshot AI	1666	EQ-Bench
38	GPT-5.4 mini	OpenAI	1665	EQ-Bench
39	GPT-5	OpenAI	1627	EQ-Bench
40	Inkling	Thinking Machines Lab	1611	EQ-Bench
41	Gemini 3.6 Flash	Google	1604	EQ-Bench
42	GLM-5	Z.ai (Zhipu)	1601	EQ-Bench
43	GLM-5.1	Z.ai (Zhipu)	1592	EQ-Bench
44	Claude Opus 4	Anthropic	1580	EQ-Bench
45	Grok 4.5	xAI	1579	EQ-Bench
46	Kimi K2.5	Moonshot AI	1579	EQ-Bench
47	Grok 4.20 (Non-Reasoning)	xAI	1574	EQ-Bench
48	DeepSeek V4 Flash	DeepSeek	1559	EQ-Bench
49	Gemini 3.5 Flash Lite	Google	1559	EQ-Bench
50	DeepSeek V4 Pro	DeepSeek	1553	EQ-Bench
51	DeepSeek V4.1 Flash	DeepSeek	1540	EQ-Bench
52	Gemini 3 Pro	Google	1525	EQ-Bench
53	DeepSeek-V3.2-Exp	DeepSeek	1515	EQ-Bench
54	Chatgpt 4o Latest 20250326	OpenAI	1501	EQ-Bench
55	DeepSeek-R1	DeepSeek	1500	EQ-Bench
56	MiMo-V2.5-Pro	Xiaomi	1493	EQ-Bench
57	Inkling-Small	Thinking Machines Lab	1491	EQ-Bench
58	Gemini 3.1 Pro Preview	Google	1491	EQ-Bench
59	Claude Sonnet 4	Anthropic	1483	EQ-Bench
60	Qwen3.5 397B-A17B	Alibaba (Qwen)	1478	EQ-Bench
61	Mistral Medium 3.1	Mistral AI	1476	EQ-Bench
62	DeepSeek-V3	DeepSeek	1472	EQ-Bench
63	Claude 3.5 Sonnet	Anthropic	1451	EQ-Bench
64	DeepSeek-V3.1	DeepSeek	1436	EQ-Bench
65	Gemini 2.5 Pro	Google	1421	EQ-Bench
66	Hermes 4 405B	Nous Research	1421	EQ-Bench
67	GPT-4.1	OpenAI	1420	EQ-Bench
68	GLM-4.7	Z.ai (Zhipu)	1413	EQ-Bench
69	Mistral Large 3	Mistral AI	1412	EQ-Bench
70	Claude 3.7 Sonnet	Anthropic	1412	EQ-Bench
71	GLM-4.6	Z.ai (Zhipu)	1411	EQ-Bench
72	Gemma 4 31B IT	Google	1368	EQ-Bench
73	Qwen3 235B-A22B	Alibaba (Qwen)	1366	EQ-Bench
74	MiniMax-M2.5	MiniMax	1361	EQ-Bench
75	GLM-4.5	Z.ai (Zhipu)	1343	EQ-Bench
76	Grok 4.1 Fast	xAI	1327	EQ-Bench
77	GPT-5 Mini	OpenAI	1313	EQ-Bench
78	Gemma 4 26B A4B IT	Google	1305	EQ-Bench
79	Nemotron 3.5 Lightning	NVIDIA	1280	EQ-Bench
80	DeepSeek-V3.2-Speciale	DeepSeek	1276	EQ-Bench
81	Gemma 3 27B	Google	1266	EQ-Bench
82	GPT-4.5	OpenAI	1258	EQ-Bench
83	QwQ-32B	Alibaba (Qwen)	1257	EQ-Bench
84	Mistral Small 3.2	Mistral AI	1255	EQ-Bench
85	Reka Flash 3	Reka	1228	EQ-Bench
86	Grok 3	xAI	1186	EQ-Bench
87	GPT-4.1 mini	OpenAI	1147	EQ-Bench
88	Claude 3.5 Haiku	Anthropic	1146	EQ-Bench
89	Command A	Cohere	1145	EQ-Bench
90	Gemini 2.5 Flash	Google	1137	EQ-Bench
91	Gemini 2.0 Flash (Feb 2025)	Google	1128	EQ-Bench
92	Gemma 3 12B	Google	1126	EQ-Bench
93	GLM-4.7-Flash	Z.ai (Zhipu)	1125	EQ-Bench
94	Gemma 3 4B	Google	1068	EQ-Bench
95	Pixtral Large	Mistral AI	988	EQ-Bench
96	Mistral Large	Mistral AI	985	EQ-Bench
97	gpt-oss-120b	OpenAI	961	EQ-Bench
98	GPT-4.1 nano	OpenAI	946	EQ-Bench
99	Mistral Nemo	Mistral AI	881	EQ-Bench
100	GPT-4o mini	OpenAI	873	EQ-Bench
101	Llama 3.1-405B	Meta	870	EQ-Bench
102	Llama 4 Maverick	Meta	860	EQ-Bench
103	Gemma 2 9B	Google	841	EQ-Bench
104	Llama 3.1-70B	Meta	784	EQ-Bench
105	Llama 4 Scout	Meta	783	EQ-Bench
106	Mistral Small 3.1	Mistral AI	761	EQ-Bench
107	GPT-4	OpenAI	752	EQ-Bench
108	Claude 3 Haiku	Anthropic	717	EQ-Bench
109	Llama 3.1-8B	Meta	713	EQ-Bench
110	Mistral Small 3	Mistral AI	707	EQ-Bench
111	GPT-5 Nano	OpenAI	705	EQ-Bench
112	gpt-oss-20b	OpenAI	666	EQ-Bench
113	Llama 3.2 3B	Meta	595	EQ-Bench
114	GPT-3.5-turbo	OpenAI	451	EQ-Bench
115	Llama 3.2 1B	Meta	200	EQ-Bench

Compare the leaders

Other writing & preference benchmarks

Frequently asked questions

What does EQ-Bench Creative Writing measure?

Elo ratings from pairwise comparisons of short stories written to creative prompts, judged by an LLM against a detailed rubric (v3).

Which model has the highest EQ-Bench Creative Writing score?

As of October 2026, GPT-6 Astra has the highest published EQ-Bench Creative Writing score on Noometry at 2173, out of 115 models with results.

What is the best open-weight model on EQ-Bench Creative Writing?

Kimi K3 has the highest EQ-Bench Creative Writing rating among open-weight models at 2082, ranking 5 of 115 overall.