Fantasy: entering an ancient ruin

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.

Creative Writing Hallucination

Performance Score Distribution (Top 20)

Click a model name to view its detail page.

	Score
GPT-5.4	90%
GPT-5.4 (Reasoning)	90%
GPT-5.4 (Reasoning, Low)	90%
GPT-5.1	87%
Grok 4.20 (Beta, Reasoning)	87%
GPT-5.4 Mini	86%
Qwen 3.5 397B A17B	86%
GPT-5.4 Mini (Reasoning)	85%
Claude Opus 4.6 (Reasoning)	84%
GPT-5.4 Mini (Reasoning, Low)	84%
Gemini 3.1 Pro (Preview)	83%
Qwen 3.5 35B	83%
GPT-5	82%
Qwen 3.5 9B	82%
Qwen 3.5 Flash	82%
ByteDance Seed 2.0 Mini	81%
Qwen 3.5 122B	81%
Gemini 3.1 Flash Lite (Preview)	81%
Claude Sonnet 4.6 (Reasoning)	81%
o4 Mini High	81%

	Score	Cost	Time
GPT-5.4 Mini	86%	$0.013	15.7s
GPT-5.4 Mini (Reasoning, Low)	84%	$0.014	15.9s
Gemini 3.1 Flash Lite (Preview)	81%	$0.0027	8.1s
GPT-5.4	90%	$0.042	1.3m
GPT-5.4 Mini (Reasoning)	85%	$0.020	37.9s
Mistral Small Creative	78%	$0.0007	8.7s
Qwen 3.5 Flash	82%	$0.0039	1.1m
Qwen 3.5 9B	82%	$0.0009	1.3m
GPT-5.4 (Reasoning, Low)	90%	$0.049	1.2m
Qwen3 235B A22B Instruct 2507	80%	$0.0010	41.5s
Qwen 3.5 122B	81%	$0.015	43.6s
ByteDance Seed 1.6 Flash	79%	$0.0013	29.3s
Grok 4.20 (Beta, Reasoning)	87%	$0.045	40.0s
Grok 4.1 Fast	78%	$0.0018	35.4s
Mistral Small 4	77%	$0.0013	17.2s
DeepSeek V3 (2025-03-24)	77%	$0.0013	43.2s
Rocinante 12B	78%	$0.0010	21.7s
LFM2 24B	77%	$0.0002	28.5s
Qwen 3 32B	80%	$0.0016	1.0m
GPT-5.4 Nano (Reasoning)	78%	$0.0066	26.7s

	Score	Cost	Speed	Stability
GPT-5.4 Mini	86%	$0.013	15.7s	82%
GPT-5.4	90%	$0.042	1.3m	86%
GPT-5.4 (Reasoning, Low)	90%	$0.049	1.2m	88%
Grok 4.20 (Beta, Reasoning)	87%	$0.045	40.0s	85%
GPT-5.4 Mini (Reasoning, Low)	84%	$0.014	15.9s	79%
Gemini 3.1 Flash Lite (Preview)	81%	$0.0027	8.1s	80%
GPT-5.4 Mini (Reasoning)	85%	$0.020	37.9s	80%
GPT-5.1	87%	$0.049	1.5m	86%
Qwen 3.5 Flash	82%	$0.0039	1.1m	81%
Qwen3 235B A22B Instruct 2507	80%	$0.0010	41.5s	78%
Qwen 3.5 122B	81%	$0.015	43.6s	79%
Qwen 3.5 9B	82%	$0.0009	1.3m	79%
Qwen 3.5 35B	83%	$0.019	1.2m	80%
GPT-5.4 (Reasoning)	90%	$0.074	2.2m	87%
o4 Mini High	81%	$0.025	47.7s	79%
Rocinante 12B	78%	$0.0010	21.7s	74%
Mistral Small 4	77%	$0.0013	17.2s	74%
Grok 4.1 Fast	78%	$0.0018	35.4s	74%
ByteDance Seed 1.6 Flash	79%	$0.0013	29.3s	73%
GPT-5.4 Nano (Reasoning, Low)	77%	$0.0058	22.7s	74%

Rank	Model	Avg. Cost	Avg. Time	Stability	# 1	# 2	# 3	# 4	# 5	Total
2	GPT-5.4	$0.042	1.3m	86%	94	92	91	89	86	90%
14	GPT-5.4 (Reasoning)	$0.074	2.2m	87%	94	90	90	89	89	90%
3	GPT-5.4 (Reasoning, Low)	$0.049	1.2m	88%	91	91	90	89	88	90%
8	GPT-5.1	$0.049	1.5m	86%	89	89	88	86	86	87%
4	Grok 4.20 (Beta, Reasoning)	$0.045	40.0s	85%	89	88	87	87	84	87%
1	GPT-5.4 Mini	$0.013	15.7s	82%	88	87	86	86	82	86%
24	Qwen 3.5 397B A17B	$0.010	3.1m	82%	89	86	85	84	84	86%
7	GPT-5.4 Mini (Reasoning)	$0.020	37.9s	80%	88	87	84	84	81	85%
40	Claude Opus 4.6 (Reasoning)	$0.080	1.3m	80%	88	86	85	84	79	84%
5	GPT-5.4 Mini (Reasoning, Low)	$0.014	15.9s	79%	89	86	84	83	79	84%
87	Gemini 3.1 Pro (Preview)	$0.131	1.7m	81%	85	84	83	83	82	83%
13	Qwen 3.5 35B	$0.019	1.2m	80%	85	84	84	83	78	83%
76	GPT-5	$0.064	2.2m	77%	87	84	82	82	77	82%
12	Qwen 3.5 9B	$0.0009	1.3m	79%	85	84	83	79	78	82%
9	Qwen 3.5 Flash	$0.0039	1.1m	81%	82	82	82	82	80	82%
110	ByteDance Seed 2.0 Mini	$0.0048	5.2m	73%	88	85	81	80	75	81%
11	Qwen 3.5 122B	$0.015	43.6s	79%	83	82	82	81	78	81%
6	Gemini 3.1 Flash Lite (Preview)	$0.0027	8.1s	80%	82	82	81	81	80	81%
34	Claude Sonnet 4.6 (Reasoning)	$0.044	55.0s	76%	84	83	80	78	78	81%
15	o4 Mini High	$0.025	47.7s	79%	82	82	81	80	78	81%
63	Claude Opus 4.6	$0.073	1.1m	76%	85	81	81	80	76	81%
65	ByteDance Seed 2.0 Lite	$0.014	2.6m	74%	87	83	82	79	72	80%
10	Qwen3 235B A22B Instruct 2507	$0.0010	41.5s	78%	82	81	80	80	78	80%
25	Qwen 3.5 27B	$0.016	1.3m	77%	82	81	81	81	76	80%
33	Qwen 3 32B	$0.0016	1.0m	70%	86	85	78	77	73	80%
42	Z.AI GLM 5	$0.0078	1.3m	72%	87	80	80	77	73	79%
54	Claude Opus 4.5	$0.061	46.8s	75%	82	81	80	77	74	79%
41	Claude Sonnet 4.6	$0.027	34.0s	72%	84	84	80	73	73	79%
19	ByteDance Seed 1.6 Flash	$0.0013	29.3s	73%	84	80	78	76	75	79%
18	Grok 4.1 Fast	$0.0018	35.4s	74%	82	81	78	77	74	78%
37	Mistral Large 2	$0.014	33.3s	71%	85	79	76	76	75	78%
21	Mistral Small Creative	$0.0007	8.7s	71%	83	83	79	78	68	78%
16	Rocinante 12B	$0.0010	21.7s	74%	81	80	78	76	74	78%
22	GPT-5.4 Nano	$0.0058	20.5s	73%	82	80	77	76	75	78%
56	Z.AI GLM 4.7	$0.010	1.7m	72%	83	79	77	76	76	78%
27	Writer: Palmyra X5	$0.011	20.8s	72%	84	80	79	75	73	78%
53	o4 Mini	$0.017	33.9s	68%	88	78	76	75	73	78%
23	GPT-5.4 Nano (Reasoning)	$0.0066	26.7s	74%	81	80	77	76	74	78%
43	GPT-4.1	$0.018	49.3s	73%	81	81	78	74	74	78%
38	Mistral Large	$0.014	38.7s	72%	81	80	78	77	71	77%
28	LFM2 24B	$0.0002	28.5s	72%	82	80	77	74	73	77%
17	Mistral Small 4	$0.0013	17.2s	74%	81	79	77	76	73	77%
20	GPT-5.4 Nano (Reasoning, Low)	$0.0058	22.7s	74%	80	77	77	77	74	77%
26	DeepSeek V3 (2025-03-24)	$0.0013	43.2s	74%	80	80	79	74	73	77%
114	Claude Opus 4	$0.172	1.1m	74%	80	78	78	77	72	77%
92	WizardLM 2 8x22b	$0.0024	2.6m	69%	83	80	77	75	70	77%
39	Gemini 3 Flash (Preview, Reasoning)	$0.017	38.5s	73%	80	79	77	76	73	77%
106	MoonshotAI: Kimi K2.5	$0.020	3.7m	73%	79	79	76	75	75	77%
30	Z.AI GLM 5 Turbo	$0.0071	31.2s	73%	79	77	76	75	75	77%
57	MiniMax M2.5	$0.0027	41.6s	66%	86	79	76	76	66	76%
29	Mistral Medium 3.1	$0.0042	23.7s	72%	80	79	77	74	73	76%
61	Hermes 3 405B	$0.0028	42.7s	66%	88	78	76	72	68	76%
64	Gemini 2.5 Pro	$0.032	33.0s	70%	81	80	75	73	72	76%
51	Gemini 3 Flash (Preview)	$0.0079	19.9s	67%	83	79	73	73	71	76%
71	Z.AI GLM 4.6	$0.0066	52.5s	67%	84	78	74	73	70	76%
46	Claude Haiku 4.5	$0.0097	18.3s	68%	82	78	74	73	71	76%
97	ByteDance Seed 1.6	$0.014	2.7m	72%	78	78	75	74	73	76%
31	Mistral Small 4 (Reasoning)	$0.0019	23.3s	72%	78	78	76	73	73	76%
36	Mistral Large 3	$0.0029	34.0s	72%	78	77	76	75	71	76%
67	DeepSeek V3 (2024-12-26)	$0.0021	1.1m	68%	82	76	74	74	70	75%
35	Ministral 3 14B	$0.0006	12.0s	70%	79	78	75	72	72	75%
32	Gemma 3 4B	$0.0002	18.6s	71%	78	77	75	73	72	75%
105	GPT-5.2	$0.054	1.4m	69%	79	78	73	72	71	75%
80	GPT-5 Mini	$0.010	1.1m	68%	80	79	75	72	67	75%
49	Llama 3.1 8B	$0.0002	33.0s	69%	80	78	76	71	66	74%
69	Cohere Command R+ (Aug. 2024)	$0.016	27.7s	68%	79	78	73	72	69	74%
74	Claude Sonnet 4.5	$0.032	35.4s	70%	78	75	74	72	72	74%
45	GPT-4o Mini (temp=1)	$0.0011	34.3s	71%	77	76	75	73	70	74%
112	Grok 4	$0.060	2.2m	69%	78	76	73	73	71	74%
70	Gemma 3 27B	$0.0005	1.5m	72%	76	75	74	73	72	74%
52	Grok 4 Fast	$0.0018	24.3s	69%	77	77	73	71	70	74%
58	Gemma 3 12B	$0.0005	35.1s	68%	79	75	74	73	67	74%
77	GPT-4o, May 13th (temp=1)	$0.031	12.3s	68%	78	75	73	71	70	73%
111	DeepSeek V3.1	$0.0014	2.4m	62%	84	76	71	68	67	73%
47	Gemini 2.5 Flash (Reasoning)	$0.010	20.0s	71%	75	75	73	72	71	73%
81	Stealth: Healer Alpha	$0.0000	19.9s	63%	84	72	71	70	68	73%
68	Grok 4.20 (Beta)	$0.021	19.0s	69%	76	75	73	72	70	73%
55	Qwen 2.5 72B	$0.0010	46.0s	71%	76	74	73	72	71	73%
50	Ministral 3 8B	$0.0005	7.1s	68%	78	75	73	73	67	73%
44	Gemini 2.5 Flash	$0.0045	9.4s	71%	75	74	74	71	70	73%
84	DeepSeek V3.2	$0.0012	1.1m	68%	79	74	73	70	69	73%
60	Llama 3.1 70B	$0.0012	23.2s	68%	77	76	73	69	68	73%
72	Qwen 3.5 Plus (2026-02-15)	$0.0058	31.5s	68%	78	72	72	71	69	73%
59	Ministral 8B	$0.0003	6.6s	66%	79	73	71	70	69	72%
86	Aion 2.0	$0.0064	1.3m	70%	75	73	73	73	67	72%
79	DeepSeek-V2 Chat	$0.0024	1.3m	70%	74	74	73	71	69	72%
103	Gemini 3 Pro (Preview)	$0.052	57.1s	70%	74	73	72	72	69	72%
48	Gemini 2.5 Flash Lite	$0.0008	7.1s	69%	75	73	72	71	69	72%
93	GPT-4o, Aug. 6th (temp=0)	$0.047	28.4s	70%	74	73	73	72	68	72%
75	Hermes 3 70B	$0.0010	42.9s	68%	75	74	71	70	70	72%
78	Stealth: Hunter Alpha	$0.0000	52.5s	68%	76	74	73	69	67	72%
66	Claude 3 Haiku	$0.0027	16.0s	67%	77	72	72	70	68	72%
83	Mistral NeMO	$0.0005	8.0s	63%	81	71	70	69	67	72%
98	MiniMax M2.7	$0.0048	1.0m	65%	79	72	71	70	66	72%
90	GPT-4o, May 13th (temp=0)	$0.035	11.4s	67%	76	73	71	69	68	72%
62	Ministral 3B	$0.0001	3.1s	67%	77	73	72	69	66	71%
73	GPT-4.1 Nano	$0.0007	15.3s	66%	76	74	74	72	61	71%
96	Claude Sonnet 4	$0.028	39.4s	67%	75	74	72	69	67	71%
101	Claude 3.7 Sonnet	$0.038	43.5s	68%	74	74	71	70	68	71%
85	Ministral 3 3B	$0.0003	3.0s	63%	77	76	70	67	65	71%
82	GPT-4o Mini (temp=0)	$0.0012	37.2s	67%	74	74	72	68	66	71%
89	Z.AI GLM 4.7 Flash	$0.0016	39.8s	65%	75	73	70	69	65	71%
91	GPT-4o, Aug. 6th (temp=1)	$0.017	15.8s	65%	75	72	70	68	67	70%
94	Llama 3.1 Nemotron 70B	$0.0032	27.0s	64%	77	70	70	67	64	70%
95	Z.AI GLM 4.5	$0.0057	50.2s	67%	72	71	69	69	67	70%
99	GPT-4.1 Mini	$0.0030	26.7s	63%	76	73	70	66	62	69%
117	Nemotron 3 Super	$0.0000	2.3m	57%	79	69	65	64	64	68%
88	Claude 3.5 Haiku	$0.0029	9.2s	65%	71	70	68	67	64	68%
100	Arcee AI: Trinity Mini	$0.0003	9.2s	61%	74	69	66	65	65	68%
116	Claude 3.5 Sonnet	$0.042	32.7s	56%	80	72	66	61	59	68%
108	Arcee AI: Trinity Large (Preview)	$0.0000	57.4s	64%	70	70	69	63	62	67%
104	Gemini 2.5 Flash Lite (Reasoning)	$0.0028	33.0s	64%	68	68	67	64	63	66%
109	Inception Mercury	$0.016	22.2s	64%	68	67	66	64	63	66%
102	Inception Mercury 2	$0.0028	6.7s	64%	67	66	65	65	65	65%
113	Nemotron 3 Nano	$0.0013	1.2m	62%	68	66	66	62	61	64%
107	Stealth: Aurora Alpha	$0.0000	9.5s	62%	68	65	65	63	62	64%
118	Mistral Small 3.2 24B	$0.0049	3.1m	57%	69	67	61	61	61	64%
115	GPT-5 Nano	$0.0044	1.4m	61%	64	64	64	62	59	62%
75.55%

Median	Evaluator	Top 3	Flop 3
80.0%	"Not X but Y" pattern overuse	100o4 Mini High 100Cohere Command R+ (Aug. 2024) 100Qwen 3.5 27B	0Stealth: Hunter Alpha 0GPT-5 Nano 6Gemini 2.5 Flash Lite (Reasoning)
46.1%	Adverb-first sentence starts	98GPT-5.4 97GPT-5.4 Mini 94Writer: Palmyra X5	0Llama 3.1 70B 0Llama 3.1 Nemotron 70B 0Claude 3.5 Haiku
100.0%	Adverbs in dialogue tags	100MoonshotAI: Kimi K2.5 100Stealth: Hunter Alpha 100Mistral Large 2	25Hermes 3 70B 38GPT-5 Nano 43Arcee AI: Trinity Large (Preview)
89.6%	AI-ism adverb frequency	100Grok 4.1 Fast 99GPT-5 99GPT-5.1	51Mistral Small 3.2 24B 60GPT-4.1 Nano 65Hermes 3 70B
100.0%	AI-ism character names	100Gemma 3 27B 100o4 Mini 100Z.AI GLM 4.7 Flash	88Llama 3.1 8B 92GPT-5.4 92Grok 4.1 Fast
100.0%	AI-ism location names	100Llama 3.1 8B 100Qwen 3.5 397B A17B 100Z.AI GLM 4.7 Flash	96o4 Mini High
20.3%	AI-ism word frequency	71GPT-5.4 (Reasoning) 67ByteDance Seed 2.0 Lite 64GPT-5.4 (Reasoning, Low)	0GPT-4o, May 13th (temp=0) 0GPT-4.1 Nano 0Claude 3 Haiku
100.0%	Cliché density	100GPT-5 Mini 100Claude Sonnet 4.6 (Reasoning) 100Gemma 3 12B	47Mistral Small 3.2 24B 60Nemotron 3 Nano 67Qwen 3.5 27B
47.6%	Dialogue tag variety (said vs. fancy)	100ByteDance Seed 2.0 Mini 100Gemini 3.1 Flash Lite (Preview) 100Mistral Large 3	0Hermes 3 70B 0Grok 4.20 (Beta) 0Gemini 2.5 Flash (Reasoning)
20.0%	Em-dash & semicolon overuse	100GPT-4o, May 13th (temp=0) 100Qwen 3.5 122B 100GPT-5.4 Mini	0Gemini 3 Flash (Preview, Reasoning) 0Z.AI GLM 5 0Mistral Small Creative
100.0%	Emotion telling (show vs. tell)	100GPT-5 Mini 100Claude Sonnet 4.6 (Reasoning) 100Gemma 3 12B	36GPT-4o, Aug. 6th (temp=0) 47GPT-4o, May 13th (temp=0) 55Llama 3.1 70B
91.7%	Filter word density	100Ministral 3 3B 100Qwen 3.5 Flash 100Qwen3 235B A22B Instruct 2507	1Inception Mercury 4Llama 3.1 8B 8Nemotron 3 Nano
100.0%	Gibberish response detection	100GPT-5.4 Nano (Reasoning) 100Claude Sonnet 4.5 100Claude Opus 4.6	23Llama 3.1 8B 80Rocinante 12B 91DeepSeek V3 (2025-03-24)
100.0%	Markdown formatting overuse	100Ministral 3 14B 100Qwen 2.5 72B 100MiniMax M2.7	33Ministral 8B 60Mistral Medium 3.1 71Ministral 3 8B
100.0%	Missing dialogue indicators (quotation marks)	100GPT-4o Mini (temp=1) 100Qwen 2.5 72B 100Claude Sonnet 4.6 (Reasoning)	77Qwen 3.5 35B 80Qwen 3.5 27B 80Qwen 3.5 122B
58.5%	Name drop frequency	96Claude Sonnet 4.6 96Grok 4.20 (Beta, Reasoning) 94Nemotron 3 Nano	3GPT-5.2 6Qwen 3.5 27B 15GPT-5.4 Nano (Reasoning)
49.5%	Narrator intent-glossing	100Rocinante 12B 100Gemini 3.1 Pro (Preview) 99GPT-5.4 (Reasoning)	0Z.AI GLM 4.5 1Gemini 2.5 Flash Lite (Reasoning) 1Claude 3.5 Sonnet
100.0%	Overuse of "that" (subordinate clause padding)	100Claude 3.5 Sonnet 100Qwen 3.5 Plus (2026-02-15) 100GPT-4.1 Nano	73GPT-4o, May 13th (temp=0) 77Rocinante 12B 78ByteDance Seed 2.0 Lite
98.9%	Paragraph length variance	100Gemma 3 4B 100Mistral Small Creative 100GPT-5.4 Mini (Reasoning)	41Arcee AI: Trinity Mini 51Mistral Small 3.2 24B 52Inception Mercury 2
99.0%	Passive voice overuse	100o4 Mini High 100Claude Sonnet 4.5 100Qwen 3.5 Plus (2026-02-15)	85DeepSeek V3.1 91Gemini 2.5 Pro 92ByteDance Seed 2.0 Lite
100.0%	Past progressive (was/were + -ing) overuse	100Z.AI GLM 4.7 100MoonshotAI: Kimi K2.5 100GPT-4.1	50Z.AI GLM 4.7 Flash 69Claude Haiku 4.5 76Ministral 3 14B
100.0%	Pronoun-first sentence starts	100Gemini 3.1 Pro (Preview) 100Grok 4 Fast 100Cohere Command R+ (Aug. 2024)	47Mistral NeMO 54Mistral Small 3.2 24B 74Qwen 3.5 9B
97.0%	Purple prose (modifier overload)	100GPT-4o, Aug. 6th (temp=0) 100Grok 4 Fast 100Inception Mercury 2	81Gemini 3.1 Pro (Preview) 86Gemini 2.5 Flash (Reasoning) 89Gemini 2.5 Flash Lite (Reasoning)
100.0%	Repeated phrase echo	100ByteDance Seed 1.6 100Gemini 3.1 Flash Lite (Preview) 100ByteDance Seed 2.0 Mini	—
100.0%	Sentence length variance	100GPT-5.4 Mini (Reasoning) 100Mistral Large 2 100Z.AI GLM 4.7	82Inception Mercury 83Nemotron 3 Nano 86GPT-4o, Aug. 6th (temp=1)
55.6%	Sentence opener variety	98Claude 3.5 Haiku 91GPT-4o Mini (temp=1) 90Llama 3.1 8B	30GPT-5 Nano 34Nemotron 3 Nano 37Qwen 3.5 35B
26.6%	Subject-first sentence starts	83Hermes 3 70B 77Qwen3 235B A22B Instruct 2507 77Rocinante 12B	0GPT-5 Nano 0Stealth: Aurora Alpha 0Arcee AI: Trinity Mini
20.0%	Subordinate conjunction sentence starts	71Gemma 3 4B 64DeepSeek V3.1 63ByteDance Seed 1.6 Flash	0Qwen 3.5 Plus (2026-02-15) 0Mistral Large 3 0Qwen 3.5 122B
46.3%	Technical jargon density	100Gemini 3.1 Pro (Preview) 100Qwen 3.5 122B 97Qwen 3.5 9B	0GPT-5 Nano 0Stealth: Aurora Alpha 0Grok 4
40.9%	Useless dialogue additions	100Qwen 3.5 397B A17B 100GPT-5.4 (Reasoning, Low) 100DeepSeek V3 (2025-03-24)	0Qwen 3.5 Plus (2026-02-15) 0Cohere Command R+ (Aug. 2024) 0Claude 3 Haiku

Bad Writing Habits

Fantasy: entering an ancient ruin

Performance Score Distribution (Top 20)

Price-Performance Score Distribution (Top 20)

Most Stable Models (Top 20)

Top Overall Models (Top 20)