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
Writer: Palmyra X5	90%
Claude Sonnet 4	88%
Z.AI GLM 5	88%
MoonshotAI: Kimi K2.5	86%
Claude Sonnet 4.5	88%
Hermes 3 405B	84%
Qwen 3.5 397B A17B	87%
Claude Opus 4.6	87%
Claude Opus 4	88%
GPT-5.1	86%
Minimax M2.5	85%
Claude Opus 4.5	87%
Grok 4,1 Fast	86%
Claude 3.5 Sonnet	86%
Claude Sonnet 4.6	84%
Gemini 3.1 Pro (Preview)	84%
Hermes 3 70B	83%
WizardLM 2 8x22b	83%
GPT-5	84%
Mistral Medium 3.1	84%

	Score	Cost	Time
Writer: Palmyra X5	90%	$0.013	22.8s
Z.AI GLM 5	88%	$0.012	1.8m
Claude Sonnet 4	88%	$0.045	54.2s
Hermes 3 405B	84%	$0.0054	49.2s
Minimax M2.5	85%	$0.0043	1.8m
Claude Sonnet 4.5	88%	$0.045	41.1s
Grok 4,1 Fast	86%	$0.0021	39.6s
Hermes 3 70B	83%	$0.0015	38.9s
Mistral Medium 3.1	84%	$0.0058	42.8s
Qwen 3.5 Plus (2026-02-15)	83%	$0.0071	31.1s
Ministral 3 14B	81%	$0.0012	11.9s
Claude Sonnet 4.6	84%	$0.038	37.6s
Claude 3.5 Sonnet	86%	$0.061	33.1s
Mistral Small Creative	82%	$0.0010	9.8s
DeepSeek V3 (2025-03-24)	82%	$0.0017	44.1s
Mistral Large	83%	$0.019	34.4s
ByteDance Seed 1.6 Flash	81%	$0.0015	30.8s
Mistral Large 2	81%	$0.017	30.9s
o4 Mini	83%	$0.015	26.6s
Claude Opus 4.6	87%	$0.088	1.2m

	Score	Consistency	Stability
Writer: Palmyra X5	90%	95%	85%
Qwen 3.5 397B A17B	87%	98%	85%
Claude Opus 4.6	87%	96%	83%
Z.AI GLM 5	88%	93%	82%
MoonshotAI: Kimi K2.5	86%	93%	82%
Claude Sonnet 4.5	88%	93%	82%
GPT-5.1	86%	95%	82%
Mistral Medium 3.1	84%	98%	81%
Minimax M2.5	85%	94%	81%
Gemini 3.1 Pro (Preview)	84%	96%	81%
WizardLM 2 8x22b	83%	97%	81%
Claude Sonnet 4.6	84%	95%	80%
GPT-5	84%	96%	80%
Claude Opus 4.5	87%	94%	80%
Claude Opus 4	88%	93%	80%
Grok 4,1 Fast	86%	93%	80%
Ministral 3 14B	81%	97%	80%
Claude 3.5 Sonnet	86%	93%	79%
Hermes 3 405B	84%	91%	79%
Gemini 3 Pro (Preview)	81%	98%	79%

	Score	Cost	Speed	Stability
Writer: Palmyra X5	90%	$0.013	22.8s	85%
Claude Sonnet 4.5	88%	$0.045	41.1s	82%
Grok 4,1 Fast	86%	$0.0021	39.6s	80%
Z.AI GLM 5	88%	$0.012	1.8m	82%
Claude Sonnet 4	88%	$0.045	54.2s	79%
Mistral Medium 3.1	84%	$0.0058	42.8s	81%
Claude Opus 4.6	87%	$0.088	1.2m	83%
Hermes 3 405B	84%	$0.0054	49.2s	79%
Claude Sonnet 4.6	84%	$0.038	37.6s	80%
Ministral 3 14B	81%	$0.0012	11.9s	80%
Claude 3.5 Sonnet	86%	$0.061	33.1s	79%
Minimax M2.5	85%	$0.0043	1.8m	81%
Claude Opus 4.5	87%	$0.086	54.6s	80%
Mistral Large	83%	$0.019	34.4s	79%
GPT-5.1	86%	$0.054	1.7m	82%
Hermes 3 70B	83%	$0.0015	38.9s	78%
o4 Mini	83%	$0.015	26.6s	77%
DeepSeek V3 (2025-03-24)	82%	$0.0017	44.1s	79%
Mistral Large 3	81%	$0.0048	36.1s	78%
ByteDance Seed 1.6 Flash	81%	$0.0015	30.8s	77%

Rank	Model	Avg. Cost	Avg. Time	Stability	# 1	# 2	# 3	# 4	# 5	Total
1	Writer: Palmyra X5	$0.013	22.8s	85%	94	91	90	89	87	90%
2	Claude Sonnet 4.5	$0.045	41.1s	82%	94	90	88	87	83	88%
5	Claude Sonnet 4	$0.045	54.2s	79%	96	91	90	86	78	88%
4	Z.AI GLM 5	$0.012	1.8m	82%	92	91	88	86	83	88%
76	Claude Opus 4	$0.434	2.8m	80%	94	89	86	85	85	88%
44	Qwen 3.5 397B A17B	$0.0045	6.2m	85%	88	88	87	87	86	87%
13	Claude Opus 4.5	$0.086	54.6s	80%	93	86	86	85	84	87%
7	Claude Opus 4.6	$0.088	1.2m	83%	89	88	87	85	84	87%
3	Grok 4,1 Fast	$0.0021	39.6s	80%	92	88	86	85	81	86%
52	MoonshotAI: Kimi K2.5	$0.019	5.2m	82%	90	89	88	83	81	86%
15	GPT-5.1	$0.054	1.7m	82%	90	87	86	84	82	86%
11	Claude 3.5 Sonnet	$0.061	33.1s	79%	91	85	85	85	81	86%
12	Minimax M2.5	$0.0043	1.8m	81%	89	87	86	86	80	85%
9	Claude Sonnet 4.6	$0.038	37.6s	80%	86	86	84	84	80	84%
8	Hermes 3 405B	$0.0054	49.2s	79%	88	87	87	81	77	84%
36	GPT-5	$0.059	2.8m	80%	87	84	83	83	82	84%
41	Gemini 3.1 Pro (Preview)	$0.134	2.1m	81%	87	84	84	83	81	84%
6	Mistral Medium 3.1	$0.0058	42.8s	81%	86	84	83	83	83	84%
17	o4 Mini	$0.015	26.6s	77%	90	83	82	81	81	83%
32	WizardLM 2 8x22b	$0.0042	3.0m	81%	85	84	84	81	81	83%
16	Hermes 3 70B	$0.0015	38.9s	78%	87	85	84	83	76	83%
14	Mistral Large	$0.019	34.4s	79%	87	83	83	82	79	83%
26	DeepSeek V3.2	$0.0019	1.3m	76%	88	85	82	80	79	83%
23	Qwen 3.5 Plus (2026-02-15)	$0.0071	31.1s	74%	91	86	83	77	76	83%
47	ByteDance Seed 1.6	$0.010	1.8m	73%	89	88	81	79	76	83%
25	Claude 3.5 Haiku	$0.0050	9.3s	72%	90	83	79	79	79	82%
18	DeepSeek V3 (2025-03-24)	$0.0017	44.1s	79%	85	83	81	81	80	82%
33	Rocinante 12B	$0.0015	34.1s	72%	90	85	80	77	77	82%
22	Mistral Small Creative	$0.0010	9.8s	73%	88	87	81	78	75	82%
42	Cohere Command R+ (Aug. 2024)	$0.028	52.8s	72%	89	85	81	80	72	82%
10	Ministral 3 14B	$0.0012	11.9s	80%	83	82	82	82	79	81%
21	Mistral Large 2	$0.017	30.9s	77%	86	83	82	79	77	81%
31	o4 Mini High	$0.026	48.1s	75%	87	81	80	79	79	81%
19	Mistral Large 3	$0.0048	36.1s	78%	84	83	82	80	78	81%
20	ByteDance Seed 1.6 Flash	$0.0015	30.8s	77%	85	82	82	79	77	81%
28	Z.AI GLM 4.7 Flash	$0.0018	1.1m	77%	85	83	83	79	75	81%
24	Gemini 2.5 Flash	$0.0065	12.1s	74%	85	85	81	77	75	81%
29	Gemini 3 Pro (Preview)	$0.058	56.3s	79%	81	81	81	81	78	81%
30	GPT-5 Mini	$0.0098	1.1m	77%	84	81	80	79	78	81%
40	Z.AI GLM 4.7	$0.012	1.7m	77%	83	81	80	79	78	80%
49	DeepSeek V3 (2024-12-26)	$0.0027	1.4m	74%	85	80	79	79	75	80%
54	Claude Haiku 4.5	$0.015	24.9s	70%	89	78	78	77	76	80%
37	Claude 3 Haiku	$0.0031	14.9s	72%	84	83	77	77	76	79%
64	Grok 4	$0.051	1.6m	74%	83	81	80	80	73	79%
39	Claude 3.7 Sonnet	$0.050	47.0s	77%	82	81	80	77	76	79%
62	GPT-5.2	$0.056	1.5m	75%	83	80	78	78	77	79%
34	Gemini 2.5 Pro	$0.037	35.2s	77%	81	80	79	78	77	79%
38	GPT-4.1	$0.020	38.4s	75%	81	80	79	79	74	79%
27	Grok 4 Fast	$0.0020	19.3s	76%	81	80	78	78	76	79%
65	DeepSeek V3.1	$0.0024	2.3m	74%	83	80	79	78	73	79%
50	Z.AI GLM 4.5	$0.0061	35.2s	72%	83	81	78	78	72	78%
43	Z.AI GLM 4.6	$0.0097	46.5s	74%	82	80	78	76	76	78%
35	GPT-4o, May 13th (temp=1)	$0.042	14.6s	76%	80	80	78	77	76	78%
46	GPT-4o, Aug. 6th (temp=1)	$0.020	16.0s	73%	83	80	78	76	73	78%
55	Ministral 8B	$0.0007	20.3s	71%	86	79	78	76	71	78%
48	Gemma 3 12B	$0.0004	45.9s	74%	82	79	78	75	74	78%
63	Gemma 3 27B	$0.0010	57.7s	71%	83	81	78	73	73	77%
45	Mistral NeMO	$0.0008	7.5s	72%	82	82	80	73	70	77%
53	Ministral 3 3B	$0.0006	6.6s	72%	81	79	78	76	70	77%
51	Qwen 2.5 72B	$0.0012	30.2s	74%	78	77	76	75	74	76%
57	GPT-4.1 Mini	$0.0034	20.9s	72%	80	79	76	74	73	76%
68	DeepSeek-V2 Chat	$0.0026	52.7s	68%	85	76	74	73	73	76%
56	GPT-4o, May 13th (temp=0)	$0.041	11.2s	74%	78	78	77	75	73	76%
58	Gemini 3 Flash (Preview)	$0.0089	20.4s	72%	79	78	76	73	73	76%
61	Arcee AI: Trinity Mini	$0.0004	9.1s	71%	81	76	75	75	72	76%
67	Llama 3.1 8B	$0.0007	58.2s	71%	80	79	78	72	69	76%
60	GPT-4o, Aug. 6th (temp=0)	$0.021	12.5s	73%	78	76	75	74	74	75%
59	Ministral 3B	$0.0002	7.0s	72%	77	76	75	74	71	75%
66	Gemini 2.5 Flash Lite	$0.0012	10.8s	70%	80	76	74	73	70	74%
73	Ministral 3 8B	$0.0012	22.3s	66%	80	79	72	71	69	74%
69	GPT-4o Mini (temp=0)	$0.0016	32.9s	69%	77	73	72	72	72	73%
70	GPT-4o Mini (temp=1)	$0.0014	39.7s	70%	77	75	74	72	69	73%
75	Arcee AI: Trinity Large (Preview)	$0.0000	36.3s	64%	80	78	71	69	67	73%
72	Llama 3.1 70B	$0.0025	16.4s	68%	77	76	76	68	65	72%
71	Gemma 3 4B	$0.0003	21.4s	69%	75	73	72	71	70	72%
78	GPT-5 Nano	$0.0042	1.5m	66%	78	73	71	69	67	72%
74	Stealth: Aurora Alpha	$0.0000	8.5s	67%	76	74	72	68	67	71%
79	Llama 3.1 Nemotron 70B	$0.0064	35.7s	62%	77	76	68	68	66	71%
80	Mistral Small 3.2 24B	$0.0078	6.5m	59%	81	68	68	67	63	69%
77	GPT-4.1 Nano	$0.0010	14.3s	66%	74	70	70	67	65	69%
79.87%

Median	Evaluator	Top 3	Flop 3
100.0%	"Not X but Y" pattern overuse	100Mistral NeMO 100o4 Mini High 100GPT-4o, Aug. 6th (temp=0)	45GPT-5 Nano 65Mistral Large 2 71Claude Haiku 4.5
48.1%	Adverb-first sentence starts	97Writer: Palmyra X5 96Gemma 3 12B 93Claude Opus 4	0Llama 3.1 70B 0Z.AI GLM 4.5 0Gemini 3.1 Pro (Preview)
98.9%	Adverbs in dialogue tags	100Claude 3.5 Sonnet 100Qwen 3.5 Plus (2026-02-15) 100Cohere Command R+ (Aug. 2024)	19GPT-4.1 Nano 40Arcee AI: Trinity Large (Preview) 45Llama 3.1 Nemotron 70B
86.3%	AI-ism adverb frequency	100Grok 4,1 Fast 99o4 Mini 98ByteDance Seed 1.6 Flash	53Gemma 3 4B 55Claude 3.5 Haiku 67Arcee AI: Trinity Large (Preview)
100.0%	AI-ism character names	100Gemini 3 Pro (Preview) 100Mistral Medium 3.1 100GPT-4o Mini (temp=0)	60Claude Opus 4 88Z.AI GLM 5 92Claude 3.5 Sonnet
100.0%	AI-ism location names	100Qwen 2.5 72B 100Cohere Command R+ (Aug. 2024) 100GPT-5 Mini	96Gemini 2.5 Pro
43.1%	AI-ism word frequency	75Claude Sonnet 4.6 72Claude Opus 4.6 72ByteDance Seed 1.6 Flash	0Gemma 3 4B 0GPT-4o Mini (temp=0) 1Qwen 2.5 72B
93.3%	Cliché density	100MoonshotAI: Kimi K2.5 100Z.AI GLM 4.7 100GPT-5.1	13Mistral Small 3.2 24B 13GPT-4o Mini (temp=0) 20GPT-4o, May 13th (temp=0)
88.4%	Dialogue tag variety (said vs. fancy)	100GPT-5 100Qwen 3.5 397B A17B 100Claude Sonnet 4	11Gemini 2.5 Flash Lite 20Stealth: Aurora Alpha 24GPT-4o, Aug. 6th (temp=1)
69.7%	Em-dash & semicolon overuse	100Qwen 3.5 397B A17B 100GPT-5 100Claude Opus 4	0Grok 4 0GPT-5 Nano 0GPT-4.1 Nano
100.0%	Emotion telling (show vs. tell)	100GPT-5 Mini 100Stealth: Aurora Alpha 100Gemma 3 12B	76Arcee AI: Trinity Large (Preview) 84GPT-4o, May 13th (temp=0) 87Mistral Small 3.2 24B
98.4%	Filter word density	100GPT-5.1 100Grok 4 100Writer: Palmyra X5	46Llama 3.1 Nemotron 70B 48Llama 3.1 8B 69Hermes 3 70B
100.0%	Gibberish response detection	100Mistral Large 2 100Gemini 3 Pro (Preview) 100DeepSeek-V2 Chat	20Llama 3.1 8B 78Rocinante 12B 80Llama 3.1 70B
100.0%	Markdown formatting overuse	100GPT-5 Mini 100Cohere Command R+ (Aug. 2024) 100Ministral 8B	79Rocinante 12B 80Llama 3.1 8B 80Llama 3.1 70B
100.0%	Missing dialogue indicators (quotation marks)	100Claude 3.5 Sonnet 100Gemma 3 12B 100GPT-4o Mini (temp=1)	80Qwen 3.5 Plus (2026-02-15) 89Mistral Small 3.2 24B 95Mistral Large 2
81.3%	Name drop frequency	100DeepSeek V3.2 100DeepSeek V3.1 100Gemini 2.5 Flash	9GPT-5.2 24Z.AI GLM 4.5 29Mistral Small Creative
84.7%	Narrator intent-glossing	100Claude 3.5 Sonnet 100Qwen 3.5 Plus (2026-02-15) 100Hermes 3 405B	26Claude Haiku 4.5 27GPT-5.2 28Llama 3.1 8B
100.0%	Overuse of "that" (subordinate clause padding)	100GPT-4.1 Nano 100Qwen 3.5 Plus (2026-02-15) 100GPT-5 Nano	49Mistral Small 3.2 24B 65Claude 3.5 Haiku 79Claude Haiku 4.5
100.0%	Paragraph length variance	100Claude Sonnet 4.5 100Qwen 2.5 72B 100Ministral 8B	64Llama 3.1 Nemotron 70B 71Stealth: Aurora Alpha 73GPT-5 Nano
99.3%	Passive voice overuse	100Claude 3.5 Sonnet 100Claude Sonnet 4.5 100Qwen 3.5 Plus (2026-02-15)	87Llama 3.1 Nemotron 70B 93Hermes 3 70B 94Gemini 2.5 Pro
100.0%	Past progressive (was/were + -ing) overuse	100DeepSeek-V2 Chat 100GPT-4o, May 13th (temp=1) 100GPT-4.1	60Llama 3.1 70B 65Z.AI GLM 4.6 69Rocinante 12B
49.1%	Pronoun-first sentence starts	99Claude 3.5 Haiku 99Mistral Small Creative 98Llama 3.1 70B	1o4 Mini High 1DeepSeek V3 (2024-12-26) 1GPT-4o Mini (temp=0)
96.4%	Purple prose (modifier overload)	100Qwen 2.5 72B 100Qwen 3.5 Plus (2026-02-15) 100MoonshotAI: Kimi K2.5	69Gemma 3 4B 82Claude 3.5 Haiku 86Gemini 3.1 Pro (Preview)
100.0%	Repeated phrase echo	100Stealth: Aurora Alpha 100Claude 3 Haiku 100GPT-4o Mini (temp=1)	—
100.0%	Sentence length variance	100Claude 3 Haiku 100GPT-5.2 100Minimax M2.5	74Mistral Small 3.2 24B 98Stealth: Aurora Alpha 99Llama 3.1 8B
59.6%	Sentence opener variety	94Claude 3.5 Haiku 92GPT-4o, Aug. 6th (temp=1) 89Grok 4,1 Fast	38GPT-5 Nano 41Gemini 3 Flash (Preview) 42Mistral Small 3.2 24B
25.6%	Subject-first sentence starts	88Rocinante 12B 79Writer: Palmyra X5 77Llama 3.1 8B	0Stealth: Aurora Alpha 1GPT-5.2 1GPT-4o Mini (temp=0)
28.5%	Subordinate conjunction sentence starts	90Cohere Command R+ (Aug. 2024) 78DeepSeek V3.2 77Gemma 3 4B	0Grok 4,1 Fast 0Ministral 3 3B 0Claude 3.5 Haiku
84.1%	Technical jargon density	100Mistral Small Creative 100Gemini 3.1 Pro (Preview) 100o4 Mini High	0GPT-4o Mini (temp=1) 2Ministral 3B 10GPT-5 Nano
40.4%	Useless dialogue additions	100Claude Sonnet 4.5 100Claude Opus 4 100Qwen 3.5 397B A17B	0GPT-4o Mini (temp=0) 0Mistral NeMO 0GPT-4o, Aug. 6th (temp=0)

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)