Thriller: chase through city streets

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
Qwen 3.5 397B A17B	91%
Writer: Palmyra X5	90%
Z.AI GLM 5	91%
Claude Sonnet 4.6	90%
GPT-5.1	90%
Claude Sonnet 4.5	90%
Claude Sonnet 4	86%
Claude Opus 4.6	89%
Claude Opus 4	86%
Claude Opus 4.5	87%
Rocinante 12B	87%
Minimax M2.5	86%
MoonshotAI: Kimi K2.5	87%
GPT-5	86%
ByteDance Seed 1.6 Flash	84%
DeepSeek V3 (2025-03-24)	84%
Hermes 3 405B	84%
Mistral Medium 3.1	86%
Hermes 3 70B	82%
o4 Mini High	83%

	Score	Cost	Time
Writer: Palmyra X5	90%	$0.011	18.7s
Z.AI GLM 5	91%	$0.0075	44.3s
Rocinante 12B	87%	$0.0015	36.4s
Claude Sonnet 4.6	90%	$0.036	40.3s
Claude Sonnet 4.5	90%	$0.041	37.3s
ByteDance Seed 1.6 Flash	84%	$0.0012	24.5s
DeepSeek V3 (2025-03-24)	84%	$0.0016	14.5s
Claude Sonnet 4	86%	$0.038	44.7s
Minimax M2.5	86%	$0.0034	1.6m
Qwen 3.5 397B A17B	91%	$0.0049	3.3m
Hermes 3 70B	82%	$0.0015	21.9s
Hermes 3 405B	84%	$0.0054	37.5s
Z.AI GLM 4.5	84%	$0.0039	28.5s
Mistral Medium 3.1	86%	$0.0059	40.6s
Arcee AI: Trinity Large (Preview)	81%	$0.0000	45.9s
o4 Mini High	83%	$0.021	44.5s
Gemma 3 27B	82%	$0.0008	51.3s
Grok 4,1 Fast	83%	$0.0021	30.9s
Claude 3.5 Haiku	83%	$0.0055	12.3s
Claude Haiku 4.5	83%	$0.014	22.0s

	Score	Cost	Speed	Stability
Writer: Palmyra X5	90%	$0.011	18.7s	87%
Z.AI GLM 5	91%	$0.0075	44.3s	86%
Qwen 3.5 397B A17B	91%	$0.0049	3.3m	87%
Claude Sonnet 4.5	90%	$0.041	37.3s	85%
Claude Sonnet 4.6	90%	$0.036	40.3s	82%
GPT-5.1	90%	$0.052	2.2m	86%
Rocinante 12B	87%	$0.0015	36.4s	80%
Mistral Medium 3.1	86%	$0.0059	40.6s	82%
Minimax M2.5	86%	$0.0034	1.6m	83%
Claude Opus 4.5	87%	$0.069	43.9s	85%
Claude Opus 4.6	89%	$0.087	1.2m	84%
DeepSeek V3 (2025-03-24)	84%	$0.0016	14.5s	80%
Claude Sonnet 4	86%	$0.038	44.7s	80%
ByteDance Seed 1.6 Flash	84%	$0.0012	24.5s	78%
Claude 3.5 Haiku	83%	$0.0055	12.3s	78%
Z.AI GLM 4.7 Flash	84%	$0.0017	1.2m	80%
Z.AI GLM 4.5	84%	$0.0039	28.5s	78%
Grok 4,1 Fast	83%	$0.0021	30.9s	78%
Hermes 3 405B	84%	$0.0054	37.5s	77%
DeepSeek V3.2	83%	$0.0020	1.3m	80%

Rank	Model	Avg. Cost	Avg. Time	Stability	# 1	# 2	# 3	# 4	# 5	Total
3	Qwen 3.5 397B A17B	$0.0049	3.3m	87%	93	93	91	90	87	91%
2	Z.AI GLM 5	$0.0075	44.3s	86%	94	92	90	90	89	91%
1	Writer: Palmyra X5	$0.011	18.7s	87%	93	91	91	89	87	90%
4	Claude Sonnet 4.5	$0.041	37.3s	85%	94	90	89	89	88	90%
5	Claude Sonnet 4.6	$0.036	40.3s	82%	95	94	89	87	84	90%
6	GPT-5.1	$0.052	2.2m	86%	92	91	89	89	86	90%
11	Claude Opus 4.6	$0.087	1.2m	84%	94	90	89	88	86	89%
10	Claude Opus 4.5	$0.069	43.9s	85%	89	88	88	86	84	87%
7	Rocinante 12B	$0.0015	36.4s	80%	92	89	87	86	81	87%
27	MoonshotAI: Kimi K2.5	$0.026	3.4m	79%	91	90	86	85	81	87%
75	Claude Opus 4	$0.292	1.9m	80%	91	89	88	85	78	86%
9	Minimax M2.5	$0.0034	1.6m	83%	89	88	87	86	82	86%
13	Claude Sonnet 4	$0.038	44.7s	80%	91	90	89	84	77	86%
34	GPT-5	$0.068	4.0m	83%	89	86	86	85	85	86%
8	Mistral Medium 3.1	$0.0059	40.6s	82%	88	87	85	84	84	86%
19	Hermes 3 405B	$0.0054	37.5s	77%	90	89	85	80	78	84%
12	DeepSeek V3 (2025-03-24)	$0.0016	14.5s	80%	88	86	85	82	79	84%
51	Gemini 3.1 Pro (Preview)	$0.136	2.2m	81%	86	84	84	84	81	84%
32	Claude 3.5 Sonnet	$0.060	38.2s	78%	88	86	83	83	79	84%
14	ByteDance Seed 1.6 Flash	$0.0012	24.5s	78%	87	85	85	85	76	84%
17	Z.AI GLM 4.5	$0.0039	28.5s	78%	88	86	84	84	76	84%
16	Z.AI GLM 4.7 Flash	$0.0017	1.2m	80%	87	84	83	83	80	84%
25	Claude Haiku 4.5	$0.014	22.0s	77%	88	87	83	82	77	83%
41	WizardLM 2 8x22b	$0.0040	2.8m	76%	90	83	82	82	79	83%
23	o4 Mini High	$0.021	44.5s	79%	87	86	84	80	78	83%
18	Grok 4,1 Fast	$0.0021	30.9s	78%	87	85	83	82	78	83%
15	Claude 3.5 Haiku	$0.0055	12.3s	78%	86	85	82	81	80	83%
20	DeepSeek V3.2	$0.0020	1.3m	80%	84	84	82	82	80	83%
42	GPT-5.2	$0.059	1.8m	79%	85	84	83	81	79	82%
24	GPT-5 Mini	$0.010	1.0m	79%	85	83	82	82	80	82%
39	Gemini 2.5 Pro	$0.036	36.4s	76%	87	85	84	80	74	82%
22	Hermes 3 70B	$0.0015	21.9s	78%	85	85	84	78	76	82%
28	Gemma 3 27B	$0.0008	51.3s	76%	86	85	84	79	74	82%
26	Mistral Large	$0.018	33.8s	78%	84	83	83	82	76	82%
36	o4 Mini	$0.017	29.4s	75%	86	85	81	79	77	82%
59	Z.AI GLM 4.6	$0.0066	1.0m	68%	91	87	77	77	75	81%
38	Z.AI GLM 4.7	$0.011	1.3m	77%	86	83	82	78	78	81%
31	Arcee AI: Trinity Large (Preview)	$0.0000	45.9s	76%	85	85	84	76	75	81%
21	Gemini 2.5 Flash	$0.0057	10.8s	79%	82	81	81	81	79	81%
35	Gemini 2.5 Flash Lite	$0.0009	8.1s	74%	86	84	81	81	73	81%
43	DeepSeek V3.1	$0.0028	2.4m	77%	84	81	81	80	78	81%
30	Mistral Large 3	$0.0043	29.5s	76%	84	84	82	79	75	81%
44	GPT-4.1	$0.019	55.8s	75%	85	83	80	79	77	81%
37	Cohere Command R+ (Aug. 2024)	$0.025	31.5s	77%	84	81	81	80	77	80%
33	Mistral Large 2	$0.017	30.1s	77%	83	82	81	80	76	80%
29	Mistral Small Creative	$0.0009	9.6s	76%	84	83	80	78	78	80%
48	GPT-4o Mini (temp=1)	$0.0013	41.8s	73%	87	80	79	76	74	79%
53	Claude 3 Haiku	$0.0029	11.5s	70%	85	83	77	77	74	79%
46	Qwen 3.5 Plus (2026-02-15)	$0.0068	28.0s	74%	84	79	78	77	76	79%
52	DeepSeek V3 (2024-12-26)	$0.0027	1.2m	73%	83	82	77	76	76	79%
72	DeepSeek-V2 Chat	$0.0026	58.1s	66%	86	85	77	76	67	78%
45	GPT-4.1 Nano	$0.0008	14.4s	73%	82	81	78	77	73	78%
40	Grok 4 Fast	$0.0020	24.4s	77%	80	79	78	78	77	78%
47	Arcee AI: Trinity Mini	$0.0004	7.3s	73%	83	80	78	77	74	78%
70	Claude 3.7 Sonnet	$0.047	45.4s	72%	82	82	77	75	74	78%
66	GPT-4o, May 13th (temp=1)	$0.043	17.4s	72%	83	82	79	76	69	78%
61	ByteDance Seed 1.6	$0.0098	1.7m	73%	83	78	77	77	74	78%
62	Grok 4	$0.046	1.4m	75%	80	79	79	77	75	78%
49	GPT-4.1 Mini	$0.0027	17.2s	72%	83	78	77	77	72	78%
58	Ministral 3 8B	$0.0028	1.5m	73%	83	80	79	74	72	78%
50	Ministral 3 14B	$0.0012	11.9s	72%	83	82	79	73	72	78%
55	Ministral 3B	$0.0002	5.0s	71%	83	82	78	74	71	78%
60	GPT-4o, Aug. 6th (temp=1)	$0.020	16.4s	72%	82	78	76	75	74	77%
57	Llama 3.1 Nemotron 70B	$0.0061	26.6s	72%	82	79	79	73	70	77%
64	Gemma 3 4B	$0.0003	21.4s	70%	83	80	77	72	71	77%
54	GPT-4o Mini (temp=0)	$0.0014	54.5s	74%	79	78	76	76	74	77%
69	Qwen 2.5 72B	$0.0014	42.7s	70%	82	79	75	73	72	76%
56	Ministral 8B	$0.0006	8.3s	72%	81	77	76	75	72	76%
71	GPT-4o, May 13th (temp=0)	$0.051	19.6s	73%	77	77	76	74	73	76%
65	GPT-4o, Aug. 6th (temp=0)	$0.022	15.4s	73%	79	77	77	73	72	75%
76	Gemini 3 Pro (Preview)	$0.055	53.0s	69%	82	76	74	72	72	75%
74	Mistral NeMO	$0.0009	9.5s	65%	87	74	74	70	70	75%
68	Gemma 3 12B	$0.0004	35.6s	71%	80	78	77	72	70	75%
67	Ministral 3 3B	$0.0005	3.0s	70%	79	78	76	75	67	75%
63	Llama 3.1 70B	$0.0020	22.2s	72%	77	76	75	75	72	75%
77	Llama 3.1 8B	$0.0004	2.3m	68%	79	79	78	73	63	74%
73	Gemini 3 Flash (Preview)	$0.0075	17.5s	69%	79	75	73	72	71	74%
78	Stealth: Aurora Alpha	$0.0000	9.1s	67%	72	72	71	68	65	70%
79	GPT-5 Nano	$0.0045	1.7m	68%	70	70	70	68	67	69%
80	Mistral Small 3.2 24B	$0.012	11.5m	62%	72	68	67	67	60	67%
80.75%

Median	Evaluator	Top 3	Flop 3
100.0%	"Not X but Y" pattern overuse	100Qwen 3.5 397B A17B 100ByteDance Seed 1.6 Flash 100Llama 3.1 8B	43GPT-5 Nano 49Gemini 2.5 Flash 63WizardLM 2 8x22b
57.7%	Adverb-first sentence starts	100Writer: Palmyra X5 100Z.AI GLM 5 98GPT-4.1 Nano	0Llama 3.1 Nemotron 70B 4GPT-5 Mini 5Gemini 3.1 Pro (Preview)
99.7%	Adverbs in dialogue tags	100GPT-5.2 100Qwen 3.5 397B A17B 100Grok 4,1 Fast	44GPT-4o, Aug. 6th (temp=1) 56Claude 3.5 Sonnet 60GPT-4o, May 13th (temp=1)
92.8%	AI-ism adverb frequency	100Grok 4,1 Fast 100Ministral 3 3B 99o4 Mini	74Claude 3.7 Sonnet 79GPT-4.1 Nano 80Hermes 3 405B
100.0%	AI-ism character names	100Mistral Large 2 100Gemini 3 Pro (Preview) 100GPT-5.1	88Llama 3.1 8B 88Gemma 3 4B 92Grok 4,1 Fast
100.0%	AI-ism location names	100Ministral 8B 100GPT-4o Mini (temp=0) 100Qwen 3.5 Plus (2026-02-15)	—
35.0%	AI-ism word frequency	78GPT-5 76Minimax M2.5 75GPT-5 Mini	0GPT-4o, Aug. 6th (temp=1) 0GPT-4o Mini (temp=0) 0GPT-4.1 Mini
100.0%	Cliché density	100Arcee AI: Trinity Mini 100Z.AI GLM 5 100DeepSeek V3 (2025-03-24)	27Mistral NeMO 33Qwen 2.5 72B 40GPT-4o, May 13th (temp=0)
60.0%	Dialogue tag variety (said vs. fancy)	100Claude 3.5 Sonnet 100Claude Sonnet 4.6 100Claude Opus 4.6	0GPT-4o Mini (temp=1) 0Stealth: Aurora Alpha 0GPT-4o Mini (temp=0)
47.7%	Em-dash & semicolon overuse	100Claude Sonnet 4.5 100Gemini 2.5 Flash 100Z.AI GLM 4.7	0GPT-4o Mini (temp=1) 0ByteDance Seed 1.6 0Grok 4
100.0%	Emotion telling (show vs. tell)	100DeepSeek V3 (2024-12-26) 100Gemini 3.1 Pro (Preview) 100Claude Sonnet 4.6	85Mistral Small 3.2 24B 88GPT-4o, May 13th (temp=0) 91Claude 3 Haiku
96.3%	Filter word density	100Gemini 2.5 Flash 100Z.AI GLM 4.7 100Grok 4,1 Fast	39Stealth: Aurora Alpha 59Llama 3.1 8B 65Llama 3.1 Nemotron 70B
100.0%	Gibberish response detection	100Claude 3 Haiku 100GPT-5 100Z.AI GLM 4.7 Flash	20Llama 3.1 8B 96Hermes 3 70B 97Rocinante 12B
100.0%	Markdown formatting overuse	100Writer: Palmyra X5 100GPT-4o Mini (temp=1) 100Mistral Large 2	49Ministral 3B 80Ministral 3 3B 80Mistral Small Creative
100.0%	Missing dialogue indicators (quotation marks)	100Z.AI GLM 4.7 100MoonshotAI: Kimi K2.5 100Writer: Palmyra X5	48GPT-5 89Grok 4 Fast 93Z.AI GLM 4.6
72.8%	Name drop frequency	100DeepSeek V3.1 100GPT-5.1 100Gemini 2.5 Pro	7GPT-5.2 13Claude 3 Haiku 19Llama 3.1 70B
83.6%	Narrator intent-glossing	100Gemma 3 27B 100o4 Mini 100Grok 4,1 Fast	15GPT-5 Nano 46Gemini 3 Flash (Preview) 46Minimax M2.5
100.0%	Overuse of "that" (subordinate clause padding)	100Stealth: Aurora Alpha 100Gemma 3 12B 100GPT-4o Mini (temp=1)	47Mistral Small 3.2 24B 91Llama 3.1 8B 93GPT-4.1 Nano
100.0%	Paragraph length variance	100Gemma 3 12B 100Ministral 3 3B 100Gemini 3.1 Pro (Preview)	33Mistral Small 3.2 24B 43Gemini 2.5 Flash Lite 46Mistral NeMO
97.9%	Passive voice overuse	100Gemini 3.1 Pro (Preview) 100o4 Mini High 100Hermes 3 405B	88Mistral NeMO 91Gemini 2.5 Flash Lite 93GPT-4o, Aug. 6th (temp=0)
95.3%	Past progressive (was/were + -ing) overuse	100Claude Opus 4 100Llama 3.1 8B 100GPT-5 Mini	28Z.AI GLM 4.7 39Gemma 3 27B 40Mistral Small 3.2 24B
93.8%	Pronoun-first sentence starts	100Claude 3.7 Sonnet 100Grok 4,1 Fast 100Llama 3.1 70B	1Mistral Small 3.2 24B 38Mistral NeMO 47Gemini 3 Pro (Preview)
97.6%	Purple prose (modifier overload)	100DeepSeek V3 (2025-03-24) 100Grok 4 100Stealth: Aurora Alpha	82Gemini 3.1 Pro (Preview) 91Gemini 2.5 Flash 92Grok 4,1 Fast
100.0%	Repeated phrase echo	100Ministral 3 3B 100Claude 3 Haiku 100Mistral Large 2	—
100.0%	Sentence length variance	100Arcee AI: Trinity Large (Preview) 100Mistral Medium 3.1 100GPT-4o, May 13th (temp=1)	79Mistral Small 3.2 24B 91Llama 3.1 70B 92Llama 3.1 8B
56.6%	Sentence opener variety	94Grok 4,1 Fast 92Claude 3.5 Haiku 91DeepSeek V3 (2025-03-24)	25Mistral Small 3.2 24B 33Mistral NeMO 34GPT-5 Nano
41.8%	Subject-first sentence starts	96Writer: Palmyra X5 95Rocinante 12B 87Z.AI GLM 5	0Stealth: Aurora Alpha 2GPT-4o, May 13th (temp=0) 4GPT-5 Mini
27.7%	Subordinate conjunction sentence starts	78Cohere Command R+ (Aug. 2024) 76Rocinante 12B 72Gemma 3 4B	0GPT-4o, Aug. 6th (temp=0) 0Mistral Small 3.2 24B 0Ministral 3B
83.5%	Technical jargon density	100Llama 3.1 70B 100Gemini 3.1 Pro (Preview) 100o4 Mini	8GPT-5 Nano 12Claude Haiku 4.5 26Mistral Small Creative
49.7%	Useless dialogue additions	100Claude Opus 4 100Claude Sonnet 4.5 100Claude Opus 4.5	0Gemma 3 4B 0GPT-4o, Aug. 6th (temp=0) 0Ministral 3 8B

Bad Writing Habits

Thriller: chase through city streets

Performance Score Distribution (Top 20)

Price-Performance Score Distribution (Top 20)

Most Stable Models (Top 20)

Top Overall Models (Top 20)