Literary fiction: old friends reunite

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
Claude Opus 4	92%
Writer: Palmyra X5	90%
Qwen 3.5 397B A17B	87%
Mistral Large	86%
Mistral Medium 3.1	87%
Z.AI GLM 5	88%
Minimax M2.5	85%
Claude Opus 4.6	85%
Claude Sonnet 4	88%
Claude Sonnet 4.5	86%
Hermes 3 405B	85%
Claude Opus 4.5	87%
Mistral Large 2	86%
MoonshotAI: Kimi K2.5	85%
GPT-5	86%
DeepSeek V3 (2025-03-24)	85%
Grok 4,1 Fast	85%
Mistral Large 3	86%
Claude 3.5 Sonnet	84%
Claude Sonnet 4.6	86%

	Score	Cost	Time
Writer: Palmyra X5	90%	$0.014	23.9s
Mistral Medium 3.1	87%	$0.0052	38.3s
Mistral Large	86%	$0.018	33.2s
DeepSeek V3 (2025-03-24)	85%	$0.0018	38.5s
Z.AI GLM 5	88%	$0.011	1.1m
Hermes 3 405B	85%	$0.0054	35.9s
ByteDance Seed 1.6 Flash	84%	$0.0014	27.7s
Grok 4,1 Fast	85%	$0.0026	47.0s
Mistral Large 2	86%	$0.018	32.8s
Rocinante 12B	81%	$0.0018	34.3s
Mistral Large 3	86%	$0.0051	40.8s
Ministral 3 8B	80%	$0.0010	11.5s
o4 Mini	84%	$0.014	25.2s
Claude Haiku 4.5	84%	$0.017	27.9s
Mistral Small Creative	81%	$0.0009	9.0s
Claude Sonnet 4	88%	$0.043	51.6s
Z.AI GLM 4.6	82%	$0.0050	30.4s
Minimax M2.5	85%	$0.0044	2.2m
Ministral 8B	79%	$0.0006	11.8s
GPT-4.1	83%	$0.017	54.4s

	Score	Consistency	Stability
Claude Opus 4	92%	96%	89%
Writer: Palmyra X5	90%	96%	86%
Z.AI GLM 5	88%	97%	85%
Qwen 3.5 397B A17B	87%	95%	84%
o4 Mini	84%	99%	83%
WizardLM 2 8x22b	84%	97%	82%
GPT-5.2	84%	98%	82%
Minimax M2.5	85%	95%	82%
Grok 4,1 Fast	85%	96%	82%
Mistral Large 2	86%	96%	82%
GPT-5	86%	96%	82%
Claude Opus 4.6	85%	95%	81%
Claude Opus 4.5	87%	95%	81%
GPT-5.1	85%	97%	81%
MoonshotAI: Kimi K2.5	85%	95%	81%
Mistral Medium 3.1	87%	93%	81%
Claude Sonnet 4	88%	94%	81%
DeepSeek V3 (2025-03-24)	85%	95%	81%
Claude Haiku 4.5	84%	96%	81%
Claude 3.5 Sonnet	84%	95%	81%

	Score	Cost	Speed	Stability
Writer: Palmyra X5	90%	$0.014	23.9s	86%
Z.AI GLM 5	88%	$0.011	1.1m	85%
Mistral Medium 3.1	87%	$0.0052	38.3s	81%
Mistral Large 2	86%	$0.018	32.8s	82%
Grok 4,1 Fast	85%	$0.0026	47.0s	82%
DeepSeek V3 (2025-03-24)	85%	$0.0018	38.5s	81%
Mistral Large	86%	$0.018	33.2s	81%
o4 Mini	84%	$0.014	25.2s	83%
Claude Sonnet 4	88%	$0.043	51.6s	81%
ByteDance Seed 1.6 Flash	84%	$0.0014	27.7s	80%
Mistral Large 3	86%	$0.0051	40.8s	78%
Claude Haiku 4.5	84%	$0.017	27.9s	81%
Claude Sonnet 4.5	86%	$0.048	47.9s	81%
Minimax M2.5	85%	$0.0044	2.2m	82%
Hermes 3 405B	85%	$0.0054	35.9s	77%
Claude Sonnet 4.6	86%	$0.041	39.3s	79%
GPT-4.1	83%	$0.017	54.4s	80%
WizardLM 2 8x22b	84%	$0.0039	2.1m	82%
Claude Opus 4.5	87%	$0.082	51.4s	81%
Z.AI GLM 4.6	82%	$0.0050	30.4s	77%

Rank	Model	Avg. Cost	Avg. Time	Stability	# 1	# 2	# 3	# 4	# 5	Total
35	Claude Opus 4	$0.250	1.5m	89%	94	94	92	90	89	92%
1	Writer: Palmyra X5	$0.014	23.9s	86%	93	90	89	89	88	90%
9	Claude Sonnet 4	$0.043	51.6s	81%	93	89	86	86	86	88%
2	Z.AI GLM 5	$0.011	1.1m	85%	90	89	87	87	85	88%
3	Mistral Medium 3.1	$0.0052	38.3s	81%	93	88	88	87	81	87%
21	Qwen 3.5 397B A17B	$0.012	3.7m	84%	89	89	88	84	84	87%
19	Claude Opus 4.5	$0.082	51.4s	81%	90	90	86	85	83	87%
7	Mistral Large	$0.018	33.2s	81%	90	90	88	84	79	86%
13	Claude Sonnet 4.5	$0.048	47.9s	81%	90	90	86	83	82	86%
11	Mistral Large 3	$0.0051	40.8s	78%	94	86	85	84	82	86%
4	Mistral Large 2	$0.018	32.8s	82%	89	86	86	86	82	86%
16	Claude Sonnet 4.6	$0.041	39.3s	79%	90	88	85	83	82	86%
31	GPT-5	$0.062	2.5m	82%	90	86	85	84	84	86%
14	Minimax M2.5	$0.0044	2.2m	82%	89	87	87	83	81	85%
5	Grok 4,1 Fast	$0.0026	47.0s	82%	88	85	85	85	82	85%
27	Claude Opus 4.6	$0.088	1.1m	81%	88	87	86	82	81	85%
6	DeepSeek V3 (2025-03-24)	$0.0018	38.5s	81%	87	87	85	85	80	85%
48	MoonshotAI: Kimi K2.5	$0.024	4.3m	81%	87	87	85	84	80	85%
15	Hermes 3 405B	$0.0054	35.9s	77%	93	86	86	84	76	85%
28	GPT-5.1	$0.055	2.1m	81%	88	86	84	83	83	85%
22	Claude 3.5 Sonnet	$0.064	41.3s	81%	88	85	85	83	81	84%
25	GPT-5.2	$0.057	1.5m	82%	85	85	84	83	83	84%
12	Claude Haiku 4.5	$0.017	27.9s	81%	86	86	84	84	80	84%
8	o4 Mini	$0.014	25.2s	83%	84	84	84	83	83	84%
18	WizardLM 2 8x22b	$0.0039	2.1m	82%	85	85	84	82	82	84%
10	ByteDance Seed 1.6 Flash	$0.0014	27.7s	80%	87	85	84	83	79	84%
17	GPT-4.1	$0.017	54.4s	80%	86	85	84	83	79	83%
69	Gemini 3.1 Pro (Preview)	$0.148	2.3m	81%	85	84	83	82	80	83%
26	DeepSeek V3.2	$0.0021	1.7m	79%	86	85	84	82	77	83%
33	Ministral 3 14B	$0.0012	16.5s	71%	91	87	79	78	77	83%
20	Z.AI GLM 4.6	$0.0050	30.4s	77%	86	83	81	80	78	82%
29	Rocinante 12B	$0.0018	34.3s	74%	88	86	83	76	75	81%
23	Mistral Small Creative	$0.0009	9.0s	75%	87	83	81	79	76	81%
24	GPT-5 Mini	$0.010	1.0m	79%	83	82	82	81	78	81%
34	DeepSeek V3 (2024-12-26)	$0.0027	44.5s	74%	87	85	80	78	76	81%
44	o4 Mini High	$0.042	1.3m	78%	83	82	81	80	79	81%
32	ByteDance Seed 1.6	$0.0093	1.7m	79%	82	82	81	80	79	81%
39	Z.AI GLM 4.5	$0.0062	46.2s	75%	85	83	79	79	77	81%
36	Claude 3.5 Haiku	$0.0052	9.8s	73%	85	85	81	79	72	81%
43	DeepSeek-V2 Chat	$0.0027	50.8s	74%	87	81	79	78	77	80%
41	Z.AI GLM 4.7	$0.012	1.8m	79%	81	81	81	80	78	80%
47	Gemini 2.5 Pro	$0.039	36.9s	76%	83	82	81	81	73	80%
51	Llama 3.1 8B	$0.0003	2.0m	75%	85	80	79	79	77	80%
45	Ministral 3 8B	$0.0010	11.5s	71%	88	84	82	77	68	80%
57	Claude 3.7 Sonnet	$0.056	55.9s	74%	86	82	81	76	75	80%
50	Claude 3 Haiku	$0.0030	13.7s	70%	90	78	78	77	77	80%
49	Z.AI GLM 4.7 Flash	$0.0018	1.8m	76%	84	81	80	79	76	80%
66	Grok 4	$0.060	1.8m	75%	84	80	79	78	78	80%
55	Arcee AI: Trinity Large (Preview)	$0.0000	47.2s	71%	87	85	79	75	72	80%
38	Qwen 3.5 Plus (2026-02-15)	$0.0070	28.0s	75%	83	83	79	79	76	80%
30	Ministral 8B	$0.0006	11.8s	75%	82	81	80	79	73	79%
46	Gemma 3 27B	$0.0010	55.7s	75%	82	82	80	75	75	79%
40	Grok 4 Fast	$0.0022	25.7s	74%	83	81	80	77	73	79%
37	Ministral 3B	$0.0002	5.4s	74%	84	79	79	77	75	79%
52	Gemini 2.5 Flash	$0.0046	9.1s	71%	82	82	76	76	75	78%
58	GPT-4o, May 13th (temp=1)	$0.043	15.5s	72%	82	81	77	76	75	78%
63	Gemini 3 Pro (Preview)	$0.062	58.2s	75%	81	79	78	77	75	78%
42	Qwen 2.5 72B	$0.0016	38.6s	76%	80	79	78	78	75	78%
73	Hermes 3 70B	$0.0017	2.3m	68%	86	84	80	72	67	78%
54	Gemma 3 12B	$0.0004	40.3s	73%	81	80	77	76	74	78%
61	DeepSeek V3.1	$0.0023	1.9m	75%	80	80	79	75	74	78%
53	GPT-4.1 Mini	$0.0026	16.2s	72%	82	79	76	75	75	77%
65	Ministral 3 3B	$0.0019	58.4s	70%	84	79	76	75	71	77%
59	GPT-4o, May 13th (temp=0)	$0.039	10.4s	73%	81	79	78	76	72	77%
60	GPT-4o, Aug. 6th (temp=1)	$0.021	23.6s	72%	80	80	77	76	71	77%
62	Mistral NeMO	$0.0009	9.6s	69%	81	81	75	74	70	76%
72	Cohere Command R+ (Aug. 2024)	$0.023	46.6s	69%	82	75	74	74	73	76%
56	Arcee AI: Trinity Mini	$0.0004	9.5s	72%	78	78	76	74	71	76%
76	GPT-4o, Aug. 6th (temp=0)	$0.050	34.3s	68%	80	77	73	72	71	75%
67	Gemini 3 Flash (Preview)	$0.0084	17.9s	70%	80	76	74	73	71	75%
71	Llama 3.1 70B	$0.0025	35.8s	69%	81	76	76	71	68	75%
75	Llama 3.1 Nemotron 70B	$0.0065	35.8s	66%	81	77	73	72	67	74%
68	Gemini 2.5 Flash Lite	$0.0011	11.3s	70%	77	76	74	70	70	73%
64	Stealth: Aurora Alpha	$0.0000	9.4s	72%	75	74	74	72	71	73%
70	GPT-4o Mini (temp=0)	$0.0015	46.9s	72%	74	73	73	72	72	73%
78	GPT-4o Mini (temp=1)	$0.0016	42.6s	63%	81	76	70	69	67	73%
80	Mistral Small 3.2 24B	$0.010	7.0m	66%	78	75	74	73	62	72%
74	GPT-4.1 Nano	$0.0008	16.6s	67%	77	73	71	70	69	72%
79	GPT-5 Nano	$0.0042	1.5m	65%	77	71	70	69	67	71%
77	Gemma 3 4B	$0.0003	22.0s	64%	79	73	71	69	64	71%
80.59%

Median	Evaluator	Top 3	Flop 3
100.0%	"Not X but Y" pattern overuse	100Mistral NeMO 100Qwen 2.5 72B 100Mistral Small 3.2 24B	0GPT-5 Nano 77Claude 3.7 Sonnet 80Llama 3.1 Nemotron 70B
48.1%	Adverb-first sentence starts	100Mistral Large 100Writer: Palmyra X5 100Ministral 3 8B	0Claude 3 Haiku 0GPT-4o, Aug. 6th (temp=0) 0Stealth: Aurora Alpha
96.2%	Adverbs in dialogue tags	100Z.AI GLM 4.7 100Gemini 3 Flash (Preview) 100GPT-5	13GPT-4.1 Nano 43Claude Haiku 4.5 55Llama 3.1 Nemotron 70B
86.4%	AI-ism adverb frequency	100Grok 4,1 Fast 98Claude Opus 4.6 97GPT-5	63GPT-4.1 Nano 66Z.AI GLM 4.5 66Gemma 3 4B
100.0%	AI-ism character names	100Gemini 3 Flash (Preview) 100Z.AI GLM 4.7 100Arcee AI: Trinity Large (Preview)	76Claude Opus 4 80Claude Opus 4.5 84Claude 3.5 Sonnet
100.0%	AI-ism location names	100Writer: Palmyra X5 100Hermes 3 405B 100Ministral 3 14B	96Rocinante 12B
43.6%	AI-ism word frequency	86GPT-5 84ByteDance Seed 1.6 Flash 77GPT-5.1	0GPT-4o Mini (temp=0) 0GPT-4o, May 13th (temp=0) 0Stealth: Aurora Alpha
93.3%	Cliché density	100GPT-5.1 100Grok 4 Fast 100DeepSeek V3.1	13GPT-4o Mini (temp=0) 40Llama 3.1 70B 53Qwen 2.5 72B
99.4%	Dialogue tag variety (said vs. fancy)	100Claude 3.5 Sonnet 100Z.AI GLM 4.5 100WizardLM 2 8x22b	9Hermes 3 70B 18Stealth: Aurora Alpha 19GPT-4o Mini (temp=1)
79.0%	Em-dash & semicolon overuse	100GPT-5.2 100Z.AI GLM 4.7 100Qwen 3.5 Plus (2026-02-15)	0ByteDance Seed 1.6 0GPT-4.1 Nano 0GPT-4o, Aug. 6th (temp=1)
100.0%	Emotion telling (show vs. tell)	100Claude Sonnet 4.5 100Mistral Medium 3.1 100DeepSeek-V2 Chat	76Mistral Small 3.2 24B 85Llama 3.1 70B 91GPT-4o, Aug. 6th (temp=0)
100.0%	Filter word density	100GPT-4.1 100Mistral Large 2 100Z.AI GLM 4.7 Flash	64Llama 3.1 8B 67Llama 3.1 70B 76Llama 3.1 Nemotron 70B
100.0%	Gibberish response detection	100Mistral Large 100Minimax M2.5 100WizardLM 2 8x22b	23Llama 3.1 8B 80Llama 3.1 70B 89Rocinante 12B
100.0%	Markdown formatting overuse	100Z.AI GLM 4.7 Flash 100Z.AI GLM 4.7 100Mistral Large 2	80Ministral 3 3B 80Ministral 3B 80Rocinante 12B
100.0%	Missing dialogue indicators (quotation marks)	100Mistral Large 3 100Gemini 2.5 Flash Lite 100Claude Sonnet 4	60GPT-5 89GPT-5 Nano 95Rocinante 12B
49.7%	Name drop frequency	100Gemma 3 12B 97GPT-4.1 Nano 97GPT-5 Nano	0GPT-5.2 5Z.AI GLM 4.5 8GPT-4o, May 13th (temp=0)
84.0%	Narrator intent-glossing	100GPT-4o, Aug. 6th (temp=0) 100Mistral Small 3.2 24B 100DeepSeek V3 (2025-03-24)	0GPT-5 Nano 26Llama 3.1 Nemotron 70B 31Gemini 2.5 Flash Lite
100.0%	Overuse of "that" (subordinate clause padding)	100GPT-4o Mini (temp=0) 100Qwen 2.5 72B 100DeepSeek V3.1	86ByteDance Seed 1.6 86Rocinante 12B 88Claude Haiku 4.5
100.0%	Paragraph length variance	100Z.AI GLM 4.7 100DeepSeek-V2 Chat 100Qwen 3.5 397B A17B	58Mistral Small 3.2 24B 73GPT-5 Nano 78Stealth: Aurora Alpha
98.9%	Passive voice overuse	100Stealth: Aurora Alpha 100Claude Opus 4 100Grok 4 Fast	84Llama 3.1 70B 89Rocinante 12B 92Claude Sonnet 4.6
100.0%	Past progressive (was/were + -ing) overuse	100ByteDance Seed 1.6 100GPT-5.2 100WizardLM 2 8x22b	49Z.AI GLM 4.7 51Z.AI GLM 4.7 Flash 61Rocinante 12B
69.2%	Pronoun-first sentence starts	100Claude Sonnet 4.6 100GPT-5.2 100Claude 3.5 Haiku	0Gemma 3 4B 1GPT-4.1 Nano 1Z.AI GLM 4.7
97.3%	Purple prose (modifier overload)	100ByteDance Seed 1.6 Flash 100Stealth: Aurora Alpha 100Grok 4	74Mistral Small 3.2 24B 77Gemma 3 4B 80Gemma 3 27B
100.0%	Repeated phrase echo	100GPT-5 Nano 100Cohere Command R+ (Aug. 2024) 100Claude Opus 4.5	—
100.0%	Sentence length variance	100Mistral NeMO 100Grok 4 100Hermes 3 70B	84Mistral Small 3.2 24B 98GPT-4o, Aug. 6th (temp=0) 99Cohere Command R+ (Aug. 2024)
62.3%	Sentence opener variety	96Claude 3.5 Haiku 95Grok 4,1 Fast 94GPT-4o, Aug. 6th (temp=1)	37GPT-5 Nano 38Z.AI GLM 4.7 Flash 39Gemini 3 Flash (Preview)
18.5%	Subject-first sentence starts	96Writer: Palmyra X5 84Llama 3.1 8B 82Claude Opus 4	0Stealth: Aurora Alpha 0DeepSeek V3.1 0Z.AI GLM 4.7
21.2%	Subordinate conjunction sentence starts	73Llama 3.1 8B 72GPT-4.1 Nano 71Claude Haiku 4.5	0Gemini 2.5 Flash 0Claude 3.5 Sonnet 0Ministral 3 3B
81.7%	Technical jargon density	100Claude 3 Haiku 100DeepSeek V3 (2025-03-24) 100Hermes 3 405B	8GPT-5 Nano 30Gemma 3 4B 30GPT-4o Mini (temp=1)
47.7%	Useless dialogue additions	100Claude Sonnet 4.6 100Qwen 3.5 397B A17B 100Claude Sonnet 4.5	0Gemini 2.5 Pro 0Mistral Small 3.2 24B 0Stealth: Aurora Alpha

Bad Writing Habits

Literary fiction: old friends reunite

Performance Score Distribution (Top 20)

Price-Performance Score Distribution (Top 20)

Most Stable Models (Top 20)

Top Overall Models (Top 20)