Romance: separated couple reunites

Bad Writing Habits

Detects common prose quality anti-patterns in AI-generated creative writing, including passive voice, past progressive overuse, weak dialogue tags, filter words, purple prose, cliches, AI-ism words/adverbs/names, and more.

0-shot Creative writingRule following

Performance Score Distribution (Top 20)

Click a model name to view its detail page.

	Score
Mistral Large 3	84%
Mistral Small Creative	86%
GPT-5.1	87%
Claude Opus 4	86%
Claude Opus 4.6	83%
Qwen 3.5 397B A17B	85%
Mistral Medium 3.1	82%
DeepSeek V3 (2025-03-24)	84%
o4 Mini High	84%
Ministral 3 14B	84%
GPT-5	85%
Gemini 3.1 Pro (Preview)	83%
Mistral Large 2	83%
Mistral Large	85%
Z.AI GLM 5	84%
Claude 3.5 Sonnet	82%
ByteDance Seed 1.6 Flash	84%
o4 Mini	83%
Claude Sonnet 4.5	83%
Writer: Palmyra X5	83%

	Score	Cost	Time
Mistral Small Creative	86%	$0.0006	9.0s
Mistral Large 3	84%	$0.0026	28.0s
Ministral 3 14B	84%	$0.0004	9.5s
ByteDance Seed 1.6 Flash	84%	$0.0012	28.0s
Mistral Large 2	83%	$0.0097	32.4s
Mistral Medium 3.1	82%	$0.0040	36.4s
Mistral Large	85%	$0.010	27.4s
DeepSeek V3 (2025-03-24)	84%	$0.0013	1.1m
Z.AI GLM 5	84%	$0.0073	59.0s
Ministral 8B	81%	$0.0002	7.8s
o4 Mini	83%	$0.012	22.7s
o4 Mini High	84%	$0.021	39.0s
Ministral 3 8B	79%	$0.0004	8.3s
Writer: Palmyra X5	83%	$0.011	22.3s
DeepSeek V3.2	81%	$0.0011	1.3m
Gemma 3 27B	81%	$0.0005	47.9s
Minimax M2.5	82%	$0.0030	1.3m
Grok 4,1 Fast	81%	$0.0017	35.6s
GPT-4o Mini (temp=1)	78%	$0.0011	33.7s
Grok 4 Fast	79%	$0.0016	22.8s

	Score	Consistency	Stability
Mistral Small Creative	86%	96%	83%
GPT-5.1	87%	96%	83%
Claude Opus 4	86%	96%	82%
Qwen 3.5 397B A17B	85%	98%	82%
o4 Mini High	84%	98%	82%
Gemini 3.1 Pro (Preview)	83%	97%	81%
DeepSeek V3 (2025-03-24)	84%	96%	81%
Mistral Large 2	83%	96%	80%
Ministral 3 14B	84%	95%	79%
ByteDance Seed 1.6 Flash	84%	95%	79%
MoonshotAI: Kimi K2.5	82%	97%	79%
o4 Mini	83%	95%	79%
Claude Opus 4.5	80%	98%	79%
Mistral Medium 3.1	82%	94%	79%
GPT-5	85%	94%	78%
Claude 3.7 Sonnet	81%	96%	78%
Grok 4,1 Fast	81%	97%	78%
Z.AI GLM 5	84%	94%	78%
GPT-5.2	81%	96%	78%
Claude 3.5 Sonnet	82%	93%	77%

	Score	Cost	Speed	Stability
Mistral Small Creative	86%	$0.0006	9.0s	83%
Ministral 3 14B	84%	$0.0004	9.5s	79%
ByteDance Seed 1.6 Flash	84%	$0.0012	28.0s	79%
o4 Mini High	84%	$0.021	39.0s	82%
Mistral Large 2	83%	$0.0097	32.4s	80%
DeepSeek V3 (2025-03-24)	84%	$0.0013	1.1m	81%
Mistral Large	85%	$0.010	27.4s	77%
o4 Mini	83%	$0.012	22.7s	79%
Mistral Large 3	84%	$0.0026	28.0s	75%
GPT-5.1	87%	$0.047	1.5m	83%
Z.AI GLM 5	84%	$0.0073	59.0s	78%
Mistral Medium 3.1	82%	$0.0040	36.4s	79%
Writer: Palmyra X5	83%	$0.011	22.3s	76%
Grok 4,1 Fast	81%	$0.0017	35.6s	78%
Gemma 3 27B	81%	$0.0005	47.9s	77%
Grok 4 Fast	79%	$0.0016	22.8s	77%
Ministral 3 8B	79%	$0.0004	8.3s	75%
Claude 3.5 Sonnet	82%	$0.036	30.1s	77%
Claude Sonnet 4.5	83%	$0.029	33.6s	75%
Minimax M2.5	82%	$0.0030	1.3m	77%

Rank	Model	Avg. Cost	Avg. Time	Stability	# 1	# 2	# 3	# 4	# 5	Total
10	GPT-5.1	$0.047	1.5m	83%	90	87	87	85	83	87%
1	Mistral Small Creative	$0.0006	9.0s	83%	89	88	87	85	83	86%
56	Claude Opus 4	$0.175	1.2m	82%	89	87	86	86	82	86%
46	Qwen 3.5 397B A17B	$0.0068	5.4m	82%	87	85	84	84	84	85%
30	GPT-5	$0.058	1.8m	78%	89	87	83	82	81	85%
7	Mistral Large	$0.010	27.4s	77%	91	87	83	81	81	85%
9	Mistral Large 3	$0.0026	28.0s	75%	91	91	87	78	75	84%
11	Z.AI GLM 5	$0.0073	59.0s	78%	90	83	83	83	82	84%
3	ByteDance Seed 1.6 Flash	$0.0012	28.0s	79%	88	85	83	82	81	84%
4	o4 Mini High	$0.021	39.0s	82%	85	84	84	83	83	84%
2	Ministral 3 14B	$0.0004	9.5s	79%	88	85	83	81	81	84%
6	DeepSeek V3 (2025-03-24)	$0.0013	1.1m	81%	86	84	84	82	82	84%
19	Claude Sonnet 4.5	$0.029	33.6s	75%	92	83	83	80	79	83%
5	Mistral Large 2	$0.0097	32.4s	80%	85	85	83	83	80	83%
36	Claude Opus 4.6	$0.071	1.1m	77%	89	86	85	78	77	83%
13	Writer: Palmyra X5	$0.011	22.3s	76%	89	83	82	81	79	83%
49	Gemini 3.1 Pro (Preview)	$0.110	1.5m	81%	85	84	83	82	80	83%
8	o4 Mini	$0.012	22.7s	79%	87	83	83	82	79	83%
18	Claude 3.5 Sonnet	$0.036	30.1s	77%	87	85	83	80	77	82%
12	Mistral Medium 3.1	$0.0040	36.4s	79%	84	84	84	82	76	82%
20	Minimax M2.5	$0.0030	1.3m	77%	87	83	81	81	78	82%
71	MoonshotAI: Kimi K2.5	$0.022	5.7m	79%	85	81	81	81	81	82%
27	Claude Sonnet 4	$0.028	40.0s	74%	89	83	81	79	77	82%
39	Gemini 3 Pro (Preview)	$0.055	57.9s	76%	87	81	81	80	78	81%
22	DeepSeek V3.2	$0.0011	1.3m	77%	84	83	81	81	76	81%
21	Ministral 8B	$0.0002	7.8s	72%	89	84	81	80	73	81%
47	WizardLM 2 8x22b	$0.0020	2.9m	75%	86	82	80	79	79	81%
40	GPT-5.2	$0.051	1.4m	78%	83	82	81	81	77	81%
15	Gemma 3 27B	$0.0005	47.9s	77%	84	82	80	80	78	81%
25	Claude 3.7 Sonnet	$0.035	40.6s	78%	83	82	81	79	78	81%
14	Grok 4,1 Fast	$0.0017	35.6s	78%	83	81	80	80	79	81%
33	Z.AI GLM 4.6	$0.0058	58.1s	73%	87	83	79	78	76	81%
32	DeepSeek V3 (2024-12-26)	$0.0018	50.4s	72%	86	85	80	79	71	80%
23	GPT-4.1	$0.016	43.7s	77%	83	82	81	78	76	80%
38	Gemini 2.5 Pro	$0.035	36.8s	74%	85	81	80	79	74	80%
44	Claude Opus 4.5	$0.069	56.6s	79%	81	80	80	79	78	80%
24	Z.AI GLM 4.5	$0.0049	46.5s	76%	84	80	79	78	77	80%
29	Z.AI GLM 4.7	$0.0098	56.1s	76%	83	81	80	77	75	79%
16	Grok 4 Fast	$0.0016	22.8s	77%	82	80	79	78	77	79%
17	Ministral 3 8B	$0.0004	8.3s	75%	83	80	80	78	75	79%
31	GPT-5 Mini	$0.010	1.2m	77%	81	79	79	79	77	79%
26	Qwen 3.5 Plus (2026-02-15)	$0.0052	32.8s	75%	82	80	79	76	75	79%
37	Claude Haiku 4.5	$0.0098	19.9s	71%	86	80	78	77	72	79%
67	Grok 4	$0.054	2.1m	75%	81	80	78	78	75	78%
42	Rocinante 12B	$0.0015	48.0s	71%	83	82	77	76	74	78%
41	GPT-4o, May 13th (temp=1)	$0.024	12.8s	73%	82	80	80	79	70	78%
61	Hermes 3 405B	$0.0021	45.8s	66%	87	85	76	75	68	78%
28	GPT-4o Mini (temp=1)	$0.0011	33.7s	75%	81	80	80	76	73	78%
53	Cohere Command R+ (Aug. 2024)	$0.015	52.9s	72%	84	79	79	73	72	77%
74	ByteDance Seed 1.6	$0.015	2.9m	69%	83	82	76	73	73	77%
34	Ministral 3 3B	$0.0003	6.4s	73%	81	78	76	76	74	77%
35	Ministral 3B	$0.0002	14.6s	72%	82	79	78	74	73	77%
51	Hermes 3 70B	$0.0006	1.2m	72%	82	79	78	75	71	77%
65	DeepSeek-V2 Chat	$0.0017	1.1m	67%	85	79	74	73	71	77%
70	Llama 3.1 8B	$0.0001	24.9s	62%	89	82	74	70	68	77%
59	DeepSeek V3.1	$0.0020	2.1m	74%	78	77	77	75	74	76%
43	Claude 3.5 Haiku	$0.0020	7.4s	72%	80	76	76	76	71	76%
54	Claude Sonnet 4.6	$0.024	32.3s	73%	78	77	76	75	72	76%
62	Arcee AI: Trinity Large (Preview)	$0.0000	52.1s	69%	80	79	74	73	72	76%
55	Qwen 2.5 72B	$0.0006	31.5s	70%	81	76	74	74	73	75%
45	Gemini 2.5 Flash	$0.0047	10.0s	72%	78	77	76	75	71	75%
63	Llama 3.1 Nemotron 70B	$0.0024	29.5s	68%	82	78	74	72	72	75%
58	Z.AI GLM 4.7 Flash	$0.0016	59.9s	71%	79	78	74	73	72	75%
48	GPT-4o Mini (temp=0)	$0.0011	33.8s	73%	77	77	76	74	72	75%
50	GPT-4o, May 13th (temp=0)	$0.023	10.7s	73%	77	75	75	74	73	75%
69	GPT-4o, Aug. 6th (temp=1)	$0.016	29.4s	67%	81	80	74	72	69	75%
52	Mistral NeMO	$0.0003	7.4s	70%	79	77	74	73	72	75%
57	Gemini 3 Flash (Preview)	$0.0083	21.1s	71%	77	77	74	74	71	75%
64	Gemma 3 12B	$0.0003	37.0s	70%	76	75	75	74	67	74%
60	GPT-4.1 Nano	$0.0006	13.6s	69%	76	76	74	73	68	73%
68	GPT-4.1 Mini	$0.0021	15.6s	67%	78	76	73	70	67	73%
66	Stealth: Aurora Alpha	$0.0000	7.4s	67%	78	73	72	72	70	73%
72	Claude 3 Haiku	$0.0023	11.5s	66%	77	74	71	68	67	71%
79	GPT-5 Nano	$0.0047	1.5m	64%	78	72	69	68	68	71%
78	Llama 3.1 70B	$0.0008	21.5s	60%	80	76	68	67	64	71%
73	Gemma 3 4B	$0.0002	19.7s	67%	74	74	72	67	66	70%
80	Mistral Small 3.2 24B	$0.0044	3.7m	63%	79	71	70	67	64	70%
77	GPT-4o, Aug. 6th (temp=0)	$0.015	15.4s	63%	74	72	71	69	60	69%
76	Gemini 2.5 Flash Lite	$0.0009	10.3s	65%	74	72	72	69	60	69%
75	Arcee AI: Trinity Mini	$0.0003	9.4s	67%	71	70	69	67	67	69%
78.63%

Median	Evaluator	Top 3	Flop 3
100.0%	"Not X but Y" pattern overuse	100Grok 4 100Gemini 3 Pro (Preview) 100Claude Opus 4.5	21GPT-5 Nano 71Gemma 3 4B 78MoonshotAI: Kimi K2.5
41.4%	Adverb-first sentence starts	93Mistral Medium 3.1 93Minimax M2.5 91Writer: Palmyra X5	0ByteDance Seed 1.6 0GPT-4o, Aug. 6th (temp=0) 0GPT-4o Mini (temp=0)
98.1%	Adverbs in dialogue tags	100Qwen 2.5 72B 100Claude 3.5 Haiku 100Stealth: Aurora Alpha	34Llama 3.1 70B 38Llama 3.1 8B 41Hermes 3 70B
86.9%	AI-ism adverb frequency	100Gemini 3.1 Pro (Preview) 99GPT-5.2 98Grok 4,1 Fast	51Claude 3.5 Haiku 56Claude 3 Haiku 58Mistral Small 3.2 24B
100.0%	AI-ism character names	100GPT-5 Nano 100Claude 3.5 Sonnet 100Claude Sonnet 4.5	88Claude Sonnet 4 92Claude 3.7 Sonnet 92DeepSeek V3.2
100.0%	AI-ism location names	100Mistral Large 2 100GPT-4o, May 13th (temp=1) 100Arcee AI: Trinity Large (Preview)	—
36.6%	AI-ism word frequency	84Claude Sonnet 4.6 80GPT-5 78GPT-5 Mini	0Llama 3.1 Nemotron 70B 0GPT-4o Mini (temp=0) 0GPT-4o Mini (temp=1)
93.3%	Cliché density	100Gemini 2.5 Pro 100Mistral Large 3 100Stealth: Aurora Alpha	33GPT-4o, Aug. 6th (temp=0) 40Gemini 2.5 Flash Lite 47Qwen 2.5 72B
91.6%	Dialogue tag variety (said vs. fancy)	100DeepSeek V3.2 100GPT-5.1 100Claude Opus 4.5	13Hermes 3 70B 24Claude 3 Haiku 25GPT-4o, Aug. 6th (temp=1)
45.8%	Em-dash & semicolon overuse	100Qwen 3.5 397B A17B 100Gemini 3.1 Pro (Preview) 100Z.AI GLM 4.6	0GPT-4o, Aug. 6th (temp=1) 0Claude Opus 4.6 0Claude 3.5 Haiku
100.0%	Emotion telling (show vs. tell)	100Z.AI GLM 4.7 100Mistral Large 2 100Mistral Medium 3.1	85GPT-4o, May 13th (temp=0) 89Mistral Small 3.2 24B 90Mistral NeMO
98.6%	Filter word density	100Mistral Large 100DeepSeek V3 (2024-12-26) 100Claude Opus 4.6	42Llama 3.1 Nemotron 70B 61Llama 3.1 70B 66GPT-5 Nano
100.0%	Gibberish response detection	100GPT-4o, May 13th (temp=0) 100Gemini 2.5 Pro 100Ministral 3 3B	72Llama 3.1 8B 81Llama 3.1 70B 99Minimax M2.5
100.0%	Markdown formatting overuse	100GPT-4o Mini (temp=0) 100DeepSeek V3.2 100Claude Opus 4.5	80Ministral 3B 100ByteDance Seed 1.6 Flash
100.0%	Missing dialogue indicators (quotation marks)	100Mistral Small Creative 100Z.AI GLM 4.7 100Arcee AI: Trinity Large (Preview)	91o4 Mini High 95Gemini 3 Pro (Preview) 97Claude 3 Haiku
80.0%	Name drop frequency	100Z.AI GLM 4.7 Flash 100Gemini 2.5 Pro 100DeepSeek V3.1	8Ministral 3B 17GPT-5.2 24Claude 3 Haiku
84.6%	Narrator intent-glossing	100Grok 4,1 Fast 100Qwen 3.5 397B A17B 100o4 Mini	2GPT-5 Nano 33Llama 3.1 70B 36GPT-4.1 Mini
100.0%	Overuse of "that" (subordinate clause padding)	100Mistral Large 2 100DeepSeek-V2 Chat 100GPT-4.1	69Claude 3.5 Haiku 76Ministral 3 8B 77ByteDance Seed 1.6
100.0%	Paragraph length variance	100Z.AI GLM 4.6 100Gemini 2.5 Pro 100GPT-4o, May 13th (temp=0)	39Arcee AI: Trinity Mini 50Llama 3.1 Nemotron 70B 50Stealth: Aurora Alpha
99.6%	Passive voice overuse	100Gemma 3 12B 100Claude 3.5 Haiku 100Grok 4 Fast	91ByteDance Seed 1.6 92Claude 3 Haiku 93GPT-4o, Aug. 6th (temp=0)
100.0%	Past progressive (was/were + -ing) overuse	100Gemini 3.1 Pro (Preview) 100o4 Mini High 100Llama 3.1 Nemotron 70B	49Claude Sonnet 4.6 76Llama 3.1 8B 77DeepSeek V3.1
38.8%	Pronoun-first sentence starts	100Ministral 3B 99Llama 3.1 8B 97Llama 3.1 Nemotron 70B	0GPT-4o Mini (temp=0) 0GPT-5 1Gemini 2.5 Pro
97.5%	Purple prose (modifier overload)	100DeepSeek V3.2 100Grok 4 100Grok 4,1 Fast	81GPT-4.1 Nano 83Gemma 3 4B 84Llama 3.1 8B
100.0%	Repeated phrase echo	100Qwen 3.5 397B A17B 100GPT-5 Mini 100Gemma 3 12B	—
100.0%	Sentence length variance	100Gemma 3 12B 100GPT-5 Mini 100GPT-4o Mini (temp=1)	87Stealth: Aurora Alpha 97Llama 3.1 70B 98Llama 3.1 8B
56.2%	Sentence opener variety	94Claude 3.5 Haiku 88Llama 3.1 Nemotron 70B 86GPT-4o, Aug. 6th (temp=1)	35Qwen 3.5 397B A17B 39ByteDance Seed 1.6 40GPT-5 Mini
17.1%	Subject-first sentence starts	73Rocinante 12B 66Mistral Small Creative 55Claude Opus 4	0GPT-5 Mini 0Gemini 3 Flash (Preview) 0Stealth: Aurora Alpha
20.0%	Subordinate conjunction sentence starts	71Claude Sonnet 4 65Mistral Small Creative 64Qwen 3.5 397B A17B	0Mistral Small 3.2 24B 0Grok 4,1 Fast 0Gemma 3 4B
83.5%	Technical jargon density	100Qwen 3.5 397B A17B 100o4 Mini High 100GPT-4o, May 13th (temp=1)	8GPT-5 Nano 21Claude Sonnet 4.6 39Claude Haiku 4.5
40.4%	Useless dialogue additions	100GPT-5.1 100Qwen 3.5 397B A17B 100ByteDance Seed 1.6	0GPT-4.1 Nano 0GPT-4o Mini (temp=0) 0Gemini 3 Flash (Preview)

Bad Writing Habits

Romance: separated couple reunites

Performance Score Distribution (Top 20)

Price-Performance Score Distribution (Top 20)

Most Stable Models (Top 20)

Top Overall Models (Top 20)