Mystery: examining a crime scene

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 (Reasoning, Low)	93%
Claude Opus 4.6 (Reasoning)	92%
GPT-5.4	92%
GPT-5.4 (Reasoning)	91%
GPT-5.4 Mini (Reasoning)	90%
GPT-5.1	90%
Claude Sonnet 4.5	90%
GPT-5.4 Mini	89%
Writer: Palmyra X5	89%
Claude Sonnet 4.6	89%
Claude Opus 4	88%
Z.AI GLM 5	88%
Z.AI GLM 5 Turbo	88%
Qwen 3.5 397B A17B	88%
GPT-5.4 Mini (Reasoning, Low)	88%
GPT-5	87%
Claude Opus 4.6	87%
Claude Haiku 4.5	87%
Qwen3 235B A22B Instruct 2507	87%
Claude Sonnet 4.6 (Reasoning)	87%

	Score	Cost	Time
GPT-5.4 Mini	89%	$0.014	15.4s
GPT-5.4 Mini (Reasoning)	90%	$0.026	33.4s
Rocinante 12B	87%	$0.0022	45.1s
GPT-5.4 Mini (Reasoning, Low)	88%	$0.014	15.7s
Stealth: Hunter Alpha	86%	$0.0000	49.6s
Writer: Palmyra X5	89%	$0.013	21.6s
Grok 4.1 Fast	86%	$0.0017	49.2s
Z.AI GLM 5 Turbo	88%	$0.0088	33.3s
Qwen3 235B A22B Instruct 2507	87%	$0.0011	47.2s
Claude Haiku 4.5	87%	$0.015	23.7s
Mistral Large	86%	$0.018	30.5s
GPT-5.4 (Reasoning, Low)	93%	$0.054	1.2m
MiniMax M2.7	85%	$0.0043	1.3m
Grok 4.20 (Beta)	86%	$0.015	12.7s
Claude Sonnet 4.5	90%	$0.042	39.8s
GPT-5.4 Nano (Reasoning, Low)	82%	$0.0050	22.8s
Mistral Small 4	83%	$0.0017	17.4s
GPT-5.4 Nano	83%	$0.0053	18.9s
Hermes 3 405B	84%	$0.0049	25.9s
GPT-5.4	92%	$0.044	1.3m

	Score	Cost	Speed	Stability
GPT-5.4 Mini (Reasoning)	90%	$0.026	33.4s	90%
GPT-5.4 Mini	89%	$0.014	15.4s	87%
GPT-5.4 (Reasoning, Low)	93%	$0.054	1.2m	91%
Writer: Palmyra X5	89%	$0.013	21.6s	85%
GPT-5.4 Mini (Reasoning, Low)	88%	$0.014	15.7s	84%
GPT-5.4	92%	$0.044	1.3m	87%
Claude Sonnet 4.5	90%	$0.042	39.8s	85%
Z.AI GLM 5 Turbo	88%	$0.0088	33.3s	82%
Claude Haiku 4.5	87%	$0.015	23.7s	83%
Qwen3 235B A22B Instruct 2507	87%	$0.0011	47.2s	83%
GPT-5.1	90%	$0.052	1.6m	89%
Rocinante 12B	87%	$0.0022	45.1s	83%
Grok 4.1 Fast	86%	$0.0017	49.2s	83%
Grok 4.20 (Beta)	86%	$0.015	12.7s	81%
Stealth: Hunter Alpha	86%	$0.0000	49.6s	81%
Claude Opus 4.6 (Reasoning)	92%	$0.091	1.2m	88%
Claude Sonnet 4.6	89%	$0.037	40.6s	82%
GPT-5.4 Nano	83%	$0.0053	18.9s	82%
Grok 4.20 (Beta, Reasoning)	86%	$0.040	35.2s	84%
o4 Mini	84%	$0.017	30.1s	81%

Rank	Model	Avg. Cost	Avg. Time	Stability	# 1	# 2	# 3	# 4	# 5	Total
3	GPT-5.4 (Reasoning, Low)	$0.054	1.2m	91%	95	94	93	93	92	93%
16	Claude Opus 4.6 (Reasoning)	$0.091	1.2m	88%	94	94	92	89	89	92%
6	GPT-5.4	$0.044	1.3m	87%	94	94	90	90	89	92%
42	GPT-5.4 (Reasoning)	$0.090	2.6m	86%	96	92	92	90	86	91%
1	GPT-5.4 Mini (Reasoning)	$0.026	33.4s	90%	91	91	91	90	89	90%
11	GPT-5.1	$0.052	1.6m	89%	91	91	91	90	89	90%
7	Claude Sonnet 4.5	$0.042	39.8s	85%	94	91	89	88	88	90%
2	GPT-5.4 Mini	$0.014	15.4s	87%	91	90	89	89	88	89%
4	Writer: Palmyra X5	$0.013	21.6s	85%	92	90	88	88	87	89%
17	Claude Sonnet 4.6	$0.037	40.6s	82%	93	91	87	87	86	89%
112	Claude Opus 4	$0.232	2.4m	84%	92	91	89	86	85	88%
24	Z.AI GLM 5	$0.013	2.1m	82%	93	89	87	87	85	88%
8	Z.AI GLM 5 Turbo	$0.0088	33.3s	82%	93	88	87	87	85	88%
31	Qwen 3.5 397B A17B	$0.013	2.4m	82%	92	91	87	85	85	88%
5	GPT-5.4 Mini (Reasoning, Low)	$0.014	15.7s	84%	91	89	88	87	84	88%
45	GPT-5	$0.063	2.5m	86%	89	87	87	87	86	87%
38	Claude Opus 4.6	$0.077	1.2m	83%	91	89	88	85	84	87%
9	Claude Haiku 4.5	$0.015	23.7s	83%	89	89	87	86	84	87%
10	Qwen3 235B A22B Instruct 2507	$0.0011	47.2s	83%	89	89	86	86	84	87%
57	Claude Sonnet 4.6 (Reasoning)	$0.075	1.2m	79%	93	89	86	84	83	87%
12	Rocinante 12B	$0.0022	45.1s	83%	88	88	88	88	82	87%
14	Grok 4.20 (Beta)	$0.015	12.7s	81%	90	89	85	84	83	86%
15	Stealth: Hunter Alpha	$0.0000	49.6s	81%	90	89	88	87	79	86%
13	Grok 4.1 Fast	$0.0017	49.2s	83%	89	88	87	84	83	86%
19	Grok 4.20 (Beta, Reasoning)	$0.040	35.2s	84%	88	87	87	85	83	86%
23	Mistral Large	$0.018	30.5s	79%	93	88	87	82	80	86%
21	MiniMax M2.7	$0.0043	1.3m	81%	89	87	86	83	82	85%
33	WizardLM 2 8x22b	$0.0039	1.9m	80%	90	87	85	83	82	85%
34	Claude Sonnet 4	$0.035	39.0s	79%	90	88	86	81	80	85%
67	Hermes 3 70B	$0.0018	1.3m	70%	95	91	82	80	75	85%
46	Claude Opus 4.5	$0.067	40.4s	81%	88	87	85	82	81	85%
53	MoonshotAI: Kimi K2.5	$0.021	2.4m	81%	88	86	85	83	82	85%
28	MiniMax M2.5	$0.0038	1.5m	81%	88	86	84	83	82	85%
32	Hermes 3 405B	$0.0049	25.9s	75%	91	88	83	82	79	84%
59	o4 Mini High	$0.037	1.3m	78%	89	85	83	83	81	84%
20	o4 Mini	$0.017	30.1s	81%	87	85	84	83	82	84%
47	Gemma 3 27B	$0.0007	59.9s	75%	91	86	82	81	79	84%
58	ByteDance Seed 2.0 Lite	$0.011	1.9m	78%	89	84	83	81	81	84%
22	Qwen 3.5 122B	$0.015	31.5s	81%	86	84	83	83	82	84%
18	GPT-5.4 Nano	$0.0053	18.9s	82%	85	84	83	83	82	83%
27	Mistral Small 4 (Reasoning)	$0.0026	32.2s	78%	88	83	82	82	81	83%
111	Gemini 3.1 Pro (Preview)	$0.143	2.1m	80%	86	84	82	82	82	83%
30	Mistral Large 3	$0.0046	32.8s	77%	90	84	83	81	79	83%
25	GPT-5.4 Nano (Reasoning)	$0.0059	22.2s	79%	85	85	82	82	81	83%
71	GPT-5.2	$0.062	1.5m	79%	87	85	84	81	79	83%
26	Mistral Small 4	$0.0017	17.4s	78%	86	84	83	82	77	83%
52	DeepSeek V3 (2025-03-24)	$0.0022	16.1s	72%	89	89	81	77	77	82%
29	Qwen 3.5 35B	$0.013	41.6s	80%	84	83	82	81	80	82%
40	Qwen 3.5 9B	$0.0013	1.3m	79%	85	84	84	81	78	82%
39	Aion 2.0	$0.0078	1.1m	80%	84	84	82	81	81	82%
41	GPT-5.4 Nano (Reasoning, Low)	$0.0050	22.8s	75%	87	84	83	83	73	82%
35	Stealth: Healer Alpha	$0.0000	26.3s	77%	88	83	82	79	78	82%
37	Qwen 3 32B	$0.0014	44.4s	78%	86	84	83	82	77	82%
62	Mistral Large 2	$0.019	35.7s	74%	89	84	83	81	73	82%
84	Claude 3.5 Sonnet	$0.060	34.1s	72%	90	85	82	79	72	82%
66	ByteDance Seed 1.6 Flash	$0.0012	23.4s	70%	93	82	79	78	77	82%
60	DeepSeek V3.2	$0.0020	1.6m	78%	85	82	81	80	79	82%
50	GPT-5 Mini	$0.0100	1.1m	78%	84	83	83	79	77	81%
56	Z.AI GLM 4.5	$0.0054	34.0s	75%	84	84	79	79	78	81%
54	Mistral Medium 3.1	$0.0056	39.4s	76%	86	80	80	79	78	81%
48	LFM2 24B	$0.0003	30.8s	75%	86	81	80	79	78	81%
110	ByteDance Seed 2.0 Mini	$0.0041	4.1m	75%	86	82	80	78	77	81%
43	Qwen 3.5 Flash	$0.0028	41.1s	78%	83	82	81	79	78	80%
36	Mistral NeMO	$0.0009	9.6s	77%	82	82	81	80	76	80%
49	Mistral Small Creative	$0.0009	7.7s	74%	87	80	79	79	77	80%
51	Z.AI GLM 4.6	$0.0068	52.0s	78%	82	81	80	79	79	80%
76	DeepSeek V3.1	$0.0020	2.0m	75%	85	82	80	80	75	80%
65	Z.AI GLM 4.7	$0.011	1.4m	78%	82	82	81	80	76	80%
82	Arcee AI: Trinity Large (Preview)	$0.0000	38.7s	68%	92	80	79	78	70	80%
55	Qwen 3.5 Plus (2026-02-15)	$0.0075	33.3s	76%	82	82	81	80	75	80%
69	GPT-4.1	$0.021	38.6s	74%	84	82	79	78	76	80%
81	Claude 3.7 Sonnet	$0.051	53.7s	76%	82	82	79	77	77	80%
106	Qwen 3.5 27B	$0.041	2.9m	77%	82	81	81	77	77	79%
44	Ministral 3 14B	$0.0010	7.2s	76%	82	81	80	80	74	79%
78	Gemini 2.5 Pro	$0.037	33.9s	74%	84	82	80	75	75	79%
63	DeepSeek-V2 Chat	$0.0028	56.9s	76%	81	81	80	78	76	79%
74	GPT-4o, May 13th (temp=1)	$0.043	12.5s	73%	85	79	78	78	76	79%
61	Gemini 2.5 Flash	$0.0057	10.0s	74%	83	81	78	78	75	79%
103	ByteDance Seed 1.6	$0.011	2.0m	72%	84	83	80	78	69	79%
101	Gemini 3 Pro (Preview)	$0.057	52.9s	72%	84	81	78	76	75	79%
80	Cohere Command R+ (Aug. 2024)	$0.024	32.9s	73%	85	80	79	76	73	79%
92	Grok 4	$0.042	1.4m	76%	81	79	79	77	76	78%
72	Z.AI GLM 4.7 Flash	$0.0018	1.5m	76%	81	80	78	77	76	78%
75	Gemma 3 12B	$0.0004	52.6s	72%	85	82	81	72	71	78%
86	GPT-4o, May 13th (temp=0)	$0.040	10.0s	72%	85	79	78	75	73	78%
70	GPT-4o, Aug. 6th (temp=1)	$0.023	19.8s	74%	80	79	78	78	73	78%
79	Gemini 2.5 Flash (Reasoning)	$0.014	24.8s	72%	84	80	78	76	71	78%
68	GPT-4o, Aug. 6th (temp=0)	$0.019	20.5s	76%	80	79	78	76	75	78%
64	Claude 3 Haiku	$0.0028	15.0s	75%	80	79	79	74	73	77%
77	Gemini 2.5 Flash Lite	$0.0011	8.9s	69%	83	81	76	74	73	77%
73	GPT-4o Mini (temp=1)	$0.0014	38.4s	72%	80	80	76	76	73	77%
93	Gemini 3 Flash (Preview, Reasoning)	$0.011	26.6s	71%	80	78	75	75	72	76%
85	Ministral 3 3B	$0.0005	3.1s	69%	80	77	74	72	72	75%
83	Ministral 3 8B	$0.0008	7.7s	70%	80	75	75	73	72	75%
90	GPT-4.1 Mini	$0.0030	16.8s	71%	80	75	74	74	72	75%
88	Qwen 2.5 72B	$0.0010	31.4s	72%	78	76	75	73	71	75%
87	Gemini 3.1 Flash Lite (Preview)	$0.0036	9.0s	70%	78	76	74	73	72	75%
100	DeepSeek V3 (2024-12-26)	$0.0028	1.1m	71%	78	77	77	73	68	75%
91	Ministral 8B	$0.0006	6.9s	70%	79	78	77	71	67	74%
97	Llama 3.1 8B	$0.0003	39.1s	70%	79	76	74	72	70	74%
98	Llama 3.1 70B	$0.0020	17.5s	68%	80	78	76	73	65	74%
118	Mistral Small 3.2 24B	$0.0071	6.4m	70%	77	77	74	73	68	74%
96	Nemotron 3 Super	$0.0000	38.9s	71%	77	75	74	72	71	74%
105	Arcee AI: Trinity Mini	$0.0004	8.2s	66%	81	76	74	71	66	74%
104	Llama 3.1 Nemotron 70B	$0.0063	28.6s	69%	77	75	73	73	69	74%
99	Gemini 3 Flash (Preview)	$0.0079	19.4s	70%	78	75	74	71	69	73%
94	Gemma 3 4B	$0.0003	21.3s	71%	76	74	74	73	70	73%
89	Ministral 3B	$0.0002	2.6s	71%	76	74	74	72	70	73%
95	GPT-4o Mini (temp=0)	$0.0013	31.3s	72%	75	74	74	72	70	73%
102	GPT-4.1 Nano	$0.0009	16.6s	69%	75	74	74	71	66	72%
107	Claude 3.5 Haiku	$0.0052	10.6s	67%	78	72	72	71	67	72%
109	Grok 4 Fast	$0.0020	21.9s	68%	75	75	72	68	68	72%
108	Inception Mercury 2	$0.0031	6.3s	67%	75	74	70	70	68	71%
114	Nemotron 3 Nano	$0.0011	55.7s	64%	75	71	68	68	68	70%
115	Stealth: Aurora Alpha	$0.0000	9.2s	62%	77	72	69	65	62	69%
113	Gemini 2.5 Flash Lite (Reasoning)	$0.0031	24.6s	64%	72	71	68	66	65	68%
116	Inception Mercury	$0.013	19.7s	62%	75	72	69	66	60	68%
117	GPT-5 Nano	$0.0046	1.5m	65%	68	66	66	66	65	66%
80.87%

Median	Evaluator	Top 3	Flop 3
96.7%	"Not X but Y" pattern overuse	100Z.AI GLM 4.6 100Writer: Palmyra X5 100GPT-4o, May 13th (temp=0)	11Grok 4 Fast 14GPT-5 Nano 20Claude 3.5 Haiku
46.8%	Adverb-first sentence starts	100Writer: Palmyra X5 98Qwen3 235B A22B Instruct 2507 93Mistral Large	0Claude 3.5 Haiku 0Claude 3 Haiku 0Stealth: Aurora Alpha
100.0%	Adverbs in dialogue tags	100Ministral 3B 100ByteDance Seed 2.0 Lite 100Claude Sonnet 4.6 (Reasoning)	28GPT-4.1 Nano 60Claude 3.5 Haiku 64GPT-4.1 Mini
91.1%	AI-ism adverb frequency	99Inception Mercury 99Grok 4.1 Fast 99GPT-5.4 (Reasoning)	64Claude 3.5 Haiku 64GPT-4.1 Nano 73Gemini 2.5 Flash Lite (Reasoning)
100.0%	AI-ism character names	100Mistral Large 100Qwen 3.5 Plus (2026-02-15) 100Claude Opus 4.6 (Reasoning)	80Z.AI GLM 5 80Arcee AI: Trinity Large (Preview) 84Grok 4.20 (Beta)
100.0%	AI-ism location names	100GPT-4.1 Nano 100Grok 4.20 (Beta, Reasoning) 100GPT-5.4 Nano (Reasoning)	—
46.0%	AI-ism word frequency	85GPT-5.4 Mini (Reasoning) 85Claude Sonnet 4.6 84GPT-5.4 (Reasoning)	0Claude 3.5 Haiku 2Llama 3.1 Nemotron 70B 3GPT-4o, Aug. 6th (temp=0)
100.0%	Cliché density	100Ministral 3 14B 100GPT-5.4 (Reasoning, Low) 100GPT-5 Mini	27Mistral Small 3.2 24B 27Inception Mercury 33GPT-4o, May 13th (temp=0)
85.6%	Dialogue tag variety (said vs. fancy)	100Claude Opus 4.6 (Reasoning) 100Claude Sonnet 4.5 100Claude Opus 4.6	0Nemotron 3 Nano 0Stealth: Aurora Alpha 0Gemma 3 4B
80.0%	Em-dash & semicolon overuse	100Z.AI GLM 4.7 Flash 100GPT-5.2 100GPT-5	0GPT-4o Mini (temp=1) 0Grok 4 0Mistral Small 4
100.0%	Emotion telling (show vs. tell)	100Ministral 3 3B 100Qwen 3 32B 100Grok 4	93Ministral 3B 94Cohere Command R+ (Aug. 2024) 95Mistral Small 3.2 24B
97.7%	Filter word density	100o4 Mini 100Mistral Medium 3.1 100GPT-5.4 Nano	32Llama 3.1 8B 33Llama 3.1 Nemotron 70B 46Stealth: Aurora Alpha
100.0%	Gibberish response detection	100Mistral Medium 3.1 100ByteDance Seed 1.6 100GPT-5.4 Nano	49Llama 3.1 8B 59Rocinante 12B 93DeepSeek V3 (2025-03-24)
100.0%	Markdown formatting overuse	100DeepSeek-V2 Chat 100ByteDance Seed 1.6 100GPT-5.2	80Ministral 3B 80Ministral 3 3B 82Rocinante 12B
100.0%	Missing dialogue indicators (quotation marks)	100Gemini 2.5 Flash 100GPT-4.1 Mini 100GPT-5.4 Nano (Reasoning)	60GPT-5 80Qwen 3.5 27B 80o4 Mini High
47.1%	Name drop frequency	93Claude Sonnet 4.6 91ByteDance Seed 2.0 Mini 89Z.AI GLM 4.6	0GPT-4o, May 13th (temp=0) 0Claude 3 Haiku 2Llama 3.1 Nemotron 70B
86.6%	Narrator intent-glossing	100LFM2 24B 100Qwen 3.5 9B 100Qwen 3.5 122B	2GPT-5 Nano 40Claude Sonnet 4 40Inception Mercury
100.0%	Overuse of "that" (subordinate clause padding)	100Qwen 3.5 9B 100Qwen 3.5 122B 100Writer: Palmyra X5	67Llama 3.1 70B 69ByteDance Seed 2.0 Mini 80Llama 3.1 8B
100.0%	Paragraph length variance	100Gemini 3 Flash (Preview, Reasoning) 100Claude Haiku 4.5 100MiniMax M2.5	61Stealth: Aurora Alpha 74Inception Mercury 75Nemotron 3 Nano
96.0%	Passive voice overuse	100GPT-4o, Aug. 6th (temp=1) 100Qwen3 235B A22B Instruct 2507 100Gemini 3.1 Pro (Preview)	77ByteDance Seed 2.0 Lite 84Qwen 2.5 72B 85Stealth: Hunter Alpha
93.3%	Past progressive (was/were + -ing) overuse	100Grok 4.20 (Beta, Reasoning) 100Gemini 3.1 Pro (Preview) 100GPT-4o, Aug. 6th (temp=1)	17Z.AI GLM 4.7 Flash 29Z.AI GLM 4.7 38Ministral 3 8B
97.5%	Pronoun-first sentence starts	100MoonshotAI: Kimi K2.5 100Writer: Palmyra X5 100ByteDance Seed 1.6 Flash	44GPT-4.1 Nano 49Qwen 3.5 27B 55Gemini 3.1 Flash Lite (Preview)
97.6%	Purple prose (modifier overload)	100Claude Opus 4.6 100ByteDance Seed 1.6 Flash 100Mistral Small Creative	79Gemini 3 Flash (Preview, Reasoning) 81Gemini 2.5 Flash Lite (Reasoning) 82Gemma 3 4B
100.0%	Repeated phrase echo	100Gemini 3 Flash (Preview, Reasoning) 100Gemini 2.5 Pro 100MoonshotAI: Kimi K2.5	—
100.0%	Sentence length variance	100Z.AI GLM 4.5 100GPT-5 Nano 100Claude Opus 4.6 (Reasoning)	80Qwen 3.5 27B 93Nemotron 3 Nano 95Qwen 3.5 9B
66.6%	Sentence opener variety	98GPT-4o, Aug. 6th (temp=1) 96Claude 3.5 Haiku 96Hermes 3 405B	34GPT-5 Nano 35Qwen 3.5 9B 36Inception Mercury
37.8%	Subject-first sentence starts	100Writer: Palmyra X5 97Qwen3 235B A22B Instruct 2507 93GPT-5.4 (Reasoning, Low)	0Qwen 3.5 Flash 0Stealth: Aurora Alpha 0Inception Mercury
22.0%	Subordinate conjunction sentence starts	93GPT-4o Mini (temp=1) 83Mistral NeMO 73Claude Haiku 4.5	0Ministral 8B 0Gemini 3 Flash (Preview, Reasoning) 0Gemini 2.5 Flash
79.3%	Technical jargon density	100Gemini 3.1 Pro (Preview) 100Rocinante 12B 100Gemma 3 27B	1GPT-5 Nano 12Gemini 2.5 Flash Lite (Reasoning) 25MiniMax M2.7
60.3%	Useless dialogue additions	100Qwen 3.5 122B 100Claude Opus 4 100Qwen 3.5 9B	0Inception Mercury 2 0GPT-5 Nano 0Nemotron 3 Super

Bad Writing Habits

Mystery: examining a crime scene

Performance Score Distribution (Top 20)

Price-Performance Score Distribution (Top 20)

Most Stable Models (Top 20)

Top Overall Models (Top 20)