Benchmarks / 3D

UGI Writing

A maintainer-owned writing evaluation considering intelligence, style, repetition and length adherence, informed by human preferences. Maintained by DontPlanToEnd. Score: Writing ✍️.

165models

–Open Index weight

–last updated

benchmarkheavenpublisher

#	Model	Effort	Score	Cost / task	Harness
1	Gemini 3.7 Flash	max	76	–	–
2	Gemini 3.8 Flash	max	73	–	–
3	Gemini 3.5 Flash	xhigh	72	–	–
4	Gemini 3.1 Pro	high	72	–	–
5	Fable 5	max	72	–	–
6	Opus 4.5	high	70	–	–
7	Opus 5.5	max	70	–	–
8	Opus 4	medium	68	–	–
9	GPT-4.1	high	67	–	–
10	Grok 4.0709	–	67	–	–
11	Sonnet 4.5	high	67	–	–
12	Opus 4.1	high	67	–	–
13	Opus 4.6	max	66	–	–
14	o3	high	66	–	–
15	Gemini 3.6 Flash	xhigh	66	–	–
16	Gemini 3 Pro	high	66	–	–
17	Gemini 2.5 Pro	high	65	–	–
18	GPT-6 Astra	max	65	–	–
19	Gemini 3 Flash	high	65	–	–
20	Opus 4.8	max	65	–	–
21	Kimi K2.6	max	64	–	–
22	Sonnet 4	high	64	–	–
23	GPT-4o	high	63	–	–
24	Grok 4.6	max	63	–	–
25	Grok 4.20	xhigh	63	–	–
26	GPT-5	high	63	–	–
27	Kimi K2.5	max	62	–	–
28	GLM 5.2	max	61	–	–
29	o1	high	61	–	–
30	GPT-5.3	xhigh	61	–	–
31	GPT-6.1 Sol	max	61	–	–
32	GPT-5.1	high	60	–	–
33	DeepSeek V4 Pro	max	60	–	–
34	Fable 5.1	max	60	–	–
35	Grok 4.5	max	60	–	–
36	GPT-5.6 Sol	max	60	–	–
37	Sonnet 4.6	max	59	–	–
38	Grok 3	none	59	–	–
39	GPT-5.4	xhigh	58	–	–
40	Grok 4.3	xhigh	58	–	–
41	GLM 4.5	xhigh	57	–	–
42	Opus 4.7	max	57	–	–
43	GLM 5.1	max	55	–	–
44	DeepSeek V4 Flash	max	55	–	–
45	Opus 3	none	54	–	–
46	Grok 4.7	max	54	–	–
47	DeepSeek V3.2	high	54	–	–
48	DeepSeek V3.1	none	54	–	–
49	GPT-6 Sol	max	54	–	–
50	GLM 4.6	none	53	–	–
51	Sonnet 5	max	53	–	–
52	Kimi K2	medium	53	–	–
53	Haiku 4.5	xhigh	52	–	–
54	DeepSeek R1	xhigh	52	–	–
55	MiMo V2.5 Pro	xhigh	52	–	–
56	GPT-5.5	xhigh	51	–	–
57	Qwen 3.5 397B-A17B	max	50	–	–
58	GLM 5	high	49	–	–
59	DeepSeek V3.0324	–	49	–	–
60	Qwen 3 235B-A22B	xhigh	49	–	–
61	GPT-5.2025	high	48	–	–
62	GPT-5.6 Terra	max	48	–	–
63	GLM 4.5 Air	–	47	–	–
64	o4 mini	high	47	–	–
65	Qwen 3.6 Plus	max	46	–	–
66	Mistral Medium 3.5	none	45	–	–
67	Gemma 3 27B IT	medium	45	–	–
68	Gemini 2.5 Flash	medium	45	–	–
69	MiMo V2.5	xhigh	45	–	–
70	Gemma 3 27B	none	45	–	–
71	GLM 4.7	none	44	–	–
72	GPT-5.2	xhigh	44	–	–
73	DeepSeek V3	none	43	–	–
74	Gemma 4 31B IT	high	43	–	–
75	Grok 4 Fast	none	43	–	–
76	Qwen 3.6 27B	max	42	–	–
77	Qwen 3.5 27B	high	42	–	–
78	DeepSeek R1.0528	–	42	–	–
79	Grok 4.1 Fast	high	42	–	–
80	MiniMax M2.5	high	42	–	–
81	Mistral Large 3	none	42	–	–
82	MiMo V2 Flash	high	42	–	–
83	MiniMax M2.1	–	42	–	–
84	Gemma 4 26B	xhigh	42	–	–
85	Hy3	xhigh	42	–	–
86	Llama 3.3	none	41	–	–
87	GPT-5.6 Luna	max	41	–	–
88	Mistral Large 2407	–	41	–	–
89	Llama 7 0b	–	41	–	–
90	Llama 3.1 405b	none	41	–	–
91	Mistral Small 4	high	40	–	–
92	Qwen 3.5 122B-A10B	none	40	–	–
93	Mistral Medium 3.1	none	39	–	–
94	Gemma 2 7B	–	39	–	–
95	Llama 10 9b	–	39	–	–
96	Llama 33 70b	–	39	–	–
97	Qwen 3 Max	max	39	–	–
98	gpt-oss-120b	high	39	–	–
99	Mistral Large 2411	–	38	–	–
100	MiniMax M2.7	high	38	–	–
101	Qwen 3.5 35B-A3B	low	37	–	–
102	Mistral Small 3.2	none	36	–	–
103	Qwen 3.6 35B-A3B	max	36	–	–
104	GLM 4.6 V	low	35	–	–
105	Gemma 2 27B IT	–	35	–	–
106	Qwen 3 14B	none	35	–	–
107	Llama 3	–	35	–	–
108	Magistral 2509	–	34	–	–
109	Mistral Small	none	34	–	–
110	Qwen 3.5 9B	low	34	–	–
111	Qwen 2.5 32B	none	33	–	–
112	Mistral	none	33	–	–
113	Qwen 3 32B	none	33	–	–
114	Llama 3.1 70b	none	33	–	–
115	Mistral Small 2409	–	33	–	–
116	Mistral Medium 3	none	33	–	–
117	Mistral Small 3.1	none	32	–	–
118	Mistral Small 3	none	32	–	–
119	Haiku 3	medium	32	–	–
120	Qwen 3 Coder	medium	32	–	–
121	Gemma 4 12B IT	–	32	–	–
122	Devstral Small 2	none	32	–	–
123	Qwen 3 30B-A3B	none	30	–	–
124	Gemma 3 12B IT	–	30	–	–
125	Qwen 2.5 7B	–	30	–	–
126	Qwen 3.5 4B	low	30	–	–
127	Qwen 2.5 14B	–	29	–	–
128	Qwen 3	high	28	–	–
129	Qwen 2.5 72B	none	28	–	–
130	Qwen 3 8B	none	28	–	–
131	Gemma 1 2B	–	28	–	–
132	Qwen 3 16B-A3B	–	27	–	–
133	Llama 4 Scout	medium	27	–	–
134	Gemma 2 9B	–	26	–	–
135	Llama 3.3 70b	none	26	–	–
136	Llama 33	–	26	–	–
137	Qwen 3 15B-A2B	–	25	–	–
138	Gemma 2 9B IT	–	25	–	–
139	Llama 3.1 8b	medium	25	–	–
140	gpt-oss-20b	high	25	–	–
141	Gemma 3 4B IT	–	24	–	–
142	Llama 4 Maverick	medium	24	–	–
143	Qwen 3.5	none	23	–	–
144	Gemma 3 12B	none	23	–	–
145	Qwen 2.5	–	22	–	–
146	Gemma 2 2B IT	–	20	–	–
147	GLM 4	–	20	–	–
148	Llama 3 8b	none	20	–	–
149	Qwen 3.5 2B	low	19	–	–
150	Magistral Small 2509	–	19	–	–
151	Qwen 3 4B	none	19	–	–
152	Qwen 3.1	none	19	–	–
153	Qwen 2.5 Coder	none	19	–	–
154	MiniMax M2	–	17	–	–
155	Llama 3.2	–	16	–	–
156	Llama 2 70b	none	15	–	–
157	Llama 3.1	max	15	–	–
158	GLM 4.7 Flash	none	15	–	–
159	Llama 3 3b	–	14	–	–
160	Gemma 3 1B	none	13	–	–
161	Llama 3.3 8b	–	13	–	–
162	Qwen 3.0	–	13	–	–
163	Llama 3.2 1b	none	12	–	–
164	Llama 3.2 3b	none	10	–	–
165	Gemma 3 4B	none	10	–	–