Passive voice → active voice

Text Replacement

Tests deterministic text transformations: renaming characters/locations, expanding contractions, tense rewriting, POV shifts, gender swaps, combined transformations, and word avoidance. Scored by checking each expected change independently.

Text EditingHallucination

Performance Score Distribution (Top 20)

Click a model name to view its detail page.

	Score
Claude Opus 4.6	98%
Claude Opus 4.5	97%
GPT-5	97%
Claude Sonnet 4	97%
Qwen 3.5 397B A17B	97%
Grok 4.1 Fast	96%
Gemini 3 Pro (Preview)	96%
Claude Opus 4.6 (Reasoning)	96%
Claude Sonnet 4.5	96%
GPT-5.1	96%
Grok 4	96%
Gemini 3.1 Pro (Preview)	96%
Aion 2.0	96%
Z.AI GLM 5	96%
Z.AI GLM 4.7	96%
Qwen 3.5 Plus (2026-02-15)	96%
DeepSeek V3.2	96%
Claude Sonnet 4.6 (Reasoning)	96%
ByteDance Seed 1.6	95%
Gemini 2.5 Pro	95%

	Score	Cost	Time
Grok 4.1 Fast	96%	$0.0014	16.8s
Qwen 3.5 Plus (2026-02-15)	96%	$0.0021	10.9s
Gemini 2.5 Flash	95%	$0.0021	2.9s
Gemini 3 Flash (Preview)	94%	$0.0026	4.3s
Gemini 2.5 Flash Lite	90%	$0.0004	2.5s
Grok 4 Fast	93%	$0.0013	11.1s
Claude Haiku 4.5	95%	$0.0050	5.3s
DeepSeek V3.1	89%	$0.0008	37.0s
DeepSeek-V2 Chat	93%	$0.0011	22.6s
DeepSeek V3 (2024-12-26)	93%	$0.0012	23.8s
GPT-4.1 Mini	90%	$0.0015	10.8s
Mistral Large 3	91%	$0.0015	10.5s
DeepSeek V3.2	96%	$0.0008	52.0s
GPT-4.1	89%	$0.0074	7.2s
Minimax M2.5	94%	$0.0018	1.3m
Claude Sonnet 4	97%	$0.015	9.5s
Z.AI GLM 4.5	94%	$0.0055	44.8s
Claude Sonnet 4.5	96%	$0.015	7.4s
ByteDance Seed 1.6 Flash	90%	$0.0011	19.6s
Mistral Large 2	91%	$0.0061	10.7s

	Score	Cost	Speed	Stability
Qwen 3.5 Plus (2026-02-15)	96%	$0.0021	10.9s	94%
Gemini 2.5 Flash	95%	$0.0021	2.9s	92%
Grok 4.1 Fast	96%	$0.0014	16.8s	94%
Gemini 3 Flash (Preview)	94%	$0.0026	4.3s	92%
Claude Sonnet 4	97%	$0.015	9.5s	96%
Claude Haiku 4.5	95%	$0.0050	5.3s	90%
Claude Sonnet 4.5	96%	$0.015	7.4s	93%
Claude Opus 4.5	97%	$0.025	8.2s	97%
Claude Opus 4.6	98%	$0.025	8.7s	96%
Mistral Large 3	91%	$0.0015	10.5s	90%
DeepSeek-V2 Chat	93%	$0.0011	22.6s	91%
Claude Sonnet 4.6	94%	$0.015	7.5s	92%
Mistral Large 2	91%	$0.0061	10.7s	90%
Mistral Large	91%	$0.0061	10.5s	90%
DeepSeek V3 (2024-12-26)	93%	$0.0012	23.8s	90%
GPT-4.1 Mini	90%	$0.0015	10.8s	88%
Gemini 2.5 Flash (Reasoning)	94%	$0.014	21.5s	93%
Claude 3.7 Sonnet	93%	$0.015	8.4s	91%
Grok 4 Fast	93%	$0.0013	11.1s	84%
Gemini 2.5 Flash Lite (Reasoning)	95%	$0.0034	40.3s	93%

Rank	Model	Avg. Cost	Avg. Time	Stability	# 1	# 2	# 3	# 4	# 5	# 6	# 7	Total
9	Claude Opus 4.6	$0.025	8.7s	96%	98	98	98	98	98	97	97	98%
8	Claude Opus 4.5	$0.025	8.2s	97%	98	98	98	98	98	97	97	97%
45	GPT-5	$0.045	1.1m	96%	98	98	98	98	97	97	96	97%
5	Claude Sonnet 4	$0.015	9.5s	96%	97	97	97	96	96	96	96	97%
66	Qwen 3.5 397B A17B	$0.0084	3.1m	95%	98	98	98	97	97	96	95	97%
3	Grok 4.1 Fast	$0.0014	16.8s	94%	98	98	98	96	96	96	95	96%
38	Gemini 3 Pro (Preview)	$0.047	31.3s	95%	98	97	97	96	96	96	96	96%
42	Claude Opus 4.6 (Reasoning)	$0.058	25.1s	94%	98	97	97	96	96	95	95	96%
7	Claude Sonnet 4.5	$0.015	7.4s	93%	98	97	96	96	96	96	95	96%
25	GPT-5.1	$0.023	24.6s	94%	98	98	97	96	96	95	95	96%
34	Grok 4	$0.032	36.3s	93%	98	98	98	97	97	94	92	96%
79	Gemini 3.1 Pro (Preview)	$0.160	2.4m	95%	97	97	96	96	96	95	95	96%
31	Aion 2.0	$0.0054	1.1m	94%	98	98	96	96	96	95	94	96%
50	Z.AI GLM 5	$0.016	1.8m	93%	98	98	97	96	95	94	94	96%
71	Z.AI GLM 4.7	$0.013	3.6m	95%	97	97	96	96	95	95	95	96%
1	Qwen 3.5 Plus (2026-02-15)	$0.0021	10.9s	94%	96	96	96	96	96	96	93	96%
21	DeepSeek V3.2	$0.0008	52.0s	94%	97	97	96	96	96	95	94	96%
67	Claude Sonnet 4.6 (Reasoning)	$0.092	59.3s	95%	97	96	96	96	95	95	94	96%
39	ByteDance Seed 1.6	$0.0074	1.4m	93%	97	97	97	95	95	95	94	95%
46	Gemini 2.5 Pro	$0.052	40.0s	93%	98	97	96	95	94	94	93	95%
78	MoonshotAI: Kimi K2.5	$0.023	4.1m	93%	97	96	96	95	95	95	93	95%
27	Gemini 3 Flash (Preview, Reasoning)	$0.018	29.1s	93%	96	96	96	95	94	94	94	95%
6	Claude Haiku 4.5	$0.0050	5.3s	90%	99	96	95	94	94	93	93	95%
2	Gemini 2.5 Flash	$0.0021	2.9s	92%	97	96	95	95	94	94	93	95%
30	Z.AI GLM 4.6	$0.0052	58.2s	92%	98	96	95	95	94	94	93	95%
23	GPT-5 Mini	$0.0068	39.9s	93%	96	96	95	95	95	93	93	95%
20	Gemini 2.5 Flash Lite (Reasoning)	$0.0034	40.3s	93%	97	96	95	95	94	94	93	95%
37	Minimax M2.5	$0.0018	1.3m	91%	97	96	96	95	94	92	91	94%
17	Gemini 2.5 Flash (Reasoning)	$0.014	21.5s	93%	96	95	95	95	94	94	93	94%
26	Z.AI GLM 4.5	$0.0055	44.8s	93%	95	95	95	95	94	93	92	94%
4	Gemini 3 Flash (Preview)	$0.0026	4.3s	92%	95	95	95	94	94	94	91	94%
22	GPT-5.2	$0.018	14.2s	91%	95	95	95	94	93	93	91	94%
12	Claude Sonnet 4.6	$0.015	7.5s	92%	95	94	94	93	93	93	93	94%
19	Grok 4 Fast	$0.0013	11.1s	84%	97	96	96	94	94	93	81	93%
11	DeepSeek-V2 Chat	$0.0011	22.6s	91%	94	94	93	93	93	92	91	93%
15	DeepSeek V3 (2024-12-26)	$0.0012	23.8s	90%	95	94	93	93	93	91	90	93%
18	Claude 3.7 Sonnet	$0.015	8.4s	91%	94	93	93	93	93	91	91	93%
63	Claude Opus 4	$0.075	13.4s	83%	96	96	95	95	94	94	77	92%
14	Mistral Large	$0.0061	10.5s	90%	92	92	91	91	91	91	91	91%
13	Mistral Large 2	$0.0061	10.7s	90%	92	92	91	91	91	91	91	91%
10	Mistral Large 3	$0.0015	10.5s	90%	92	91	91	91	91	91	91	91%
28	ByteDance Seed 1.6 Flash	$0.0011	19.6s	85%	94	94	92	89	89	88	88	90%
16	GPT-4.1 Mini	$0.0015	10.8s	88%	92	92	91	91	91	89	87	90%
24	Gemini 2.5 Flash Lite	$0.0004	2.5s	82%	94	93	93	93	89	88	77	90%
33	GPT-4.1	$0.0074	7.2s	81%	94	94	94	93	90	81	77	89%
40	Claude 3.5 Sonnet	$0.030	13.3s	86%	91	89	89	88	88	88	88	89%
35	GPT-4o, May 13th (temp=0)	$0.015	3.6s	81%	94	91	90	90	90	89	78	89%
62	DeepSeek V3.1	$0.0008	37.0s	60%	98	97	97	96	95	95	43	89%
41	DeepSeek V3 (2025-03-24)	$0.0008	36.8s	79%	95	93	93	92	86	78	77	88%
36	Arcee AI: Trinity Large (Preview)	$0.0000	31.6s	84%	89	88	88	87	86	86	84	87%
32	Qwen 2.5 72B	$0.0004	16.0s	83%	90	89	88	88	85	84	83	87%
29	Mistral Small 3.2 24B	$0.0003	7.0s	84%	86	86	86	86	85	85	82	85%
74	o4 Mini High	$0.043	1.1m	64%	94	94	92	92	86	83	49	84%
44	GPT-4o, May 13th (temp=1)	$0.015	4.3s	76%	91	89	89	86	79	77	76	84%
47	Writer: Palmyra X5	$0.0049	17.6s	74%	90	90	87	86	86	73	72	83%
48	Llama 3.1 70B	$0.0007	13.1s	71%	91	90	87	80	80	78	74	83%
69	Z.AI GLM 4.7 Flash	$0.0027	1.8m	72%	90	89	87	84	79	78	68	82%
53	GPT-4o Mini (temp=1)	$0.0005	11.4s	69%	89	89	88	78	76	76	76	82%
68	GPT-4o, Aug. 6th (temp=0)	$0.0085	3.7s	46%	93	93	92	92	91	87	20	81%
43	Mistral Medium 3.1	$0.0018	6.2s	75%	84	83	83	79	79	78	78	81%
61	Hermes 3 70B	$0.0005	22.6s	67%	89	82	81	81	79	78	59	78%
59	Hermes 3 405B	$0.0016	36.3s	75%	79	78	78	78	76	76	75	77%
51	Gemma 3 4B	$0.0001	8.4s	73%	80	79	78	76	76	76	74	77%
49	Mistral Small Creative	$0.0003	4.1s	73%	82	78	77	77	76	76	72	77%
57	Ministral 3 14B	$0.0003	5.5s	67%	82	82	82	74	74	74	71	77%
56	Gemma 3 27B	$0.0003	23.9s	75%	77	77	77	77	77	75	75	76%
70	GPT-4o, Aug. 6th (temp=1)	$0.0084	4.6s	46%	91	89	89	86	81	78	20	76%
52	GPT-4o Mini (temp=0)	$0.0005	13.6s	76%	76	76	76	76	76	76	76	76%
54	Arcee AI: Trinity Mini	$0.0003	10.7s	74%	79	78	78	78	73	73	73	76%
55	Gemma 3 12B	$0.0001	13.5s	74%	76	76	76	76	76	76	74	75%
58	Ministral 3 8B	$0.0002	4.3s	70%	77	77	76	75	73	70	67	74%
65	Llama 3.1 Nemotron 70B	$0.0020	21.1s	63%	82	79	77	72	70	67	63	73%
64	GPT-4.1 Nano	$0.0004	5.0s	61%	84	81	78	73	66	64	60	72%
76	o4 Mini	$0.022	35.1s	54%	91	81	80	78	77	49	47	72%
60	Ministral 8B	$0.0002	4.0s	68%	76	72	72	71	71	69	68	71%
77	GPT-5 Nano	$0.0039	1.5m	62%	77	76	75	71	66	63	62	70%
81	WizardLM 2 8x22b	$0.0016	2.3m	27%	94	93	92	86	72	27	3	67%
72	Llama 3.1 8B	$0.0001	14.4s	54%	79	71	71	68	57	54	48	64%
73	Ministral 3B	$0.0001	2.8s	55%	65	62	62	61	55	52	51	58%
75	Ministral 3 3B	$0.0002	2.9s	49%	67	64	58	57	57	49	48	57%
80	Mistral NeMO	$0.0002	3.9s	22%	63	52	49	32	24	20	17	37%
82	Cohere Command R+ (Aug. 2024)	$0.0088	16.4s	24%	40	29	27	27	27	23	22	28%
83	Rocinante 12B	$0.0005	8.4s	19%	37	36	26	23	23	15	12	24%
84	Claude 3 Haiku	$0.0013	6.6s	18%	18	18	18	18	18	18	18	18%
84.20%

Median	Evaluator	Top 3	Flop 3
91.4%	Dialogue content preserved unchanged	100Claude Opus 4.6 (Reasoning) 100Z.AI GLM 4.7 100GPT-5	0Claude 3 Haiku 14Cohere Command R+ (Aug. 2024) 20Rocinante 12B
100.0%	No hallucinated or fabricated content	100GPT-5.2 100Grok 4.1 Fast 100DeepSeek-V2 Chat	14Rocinante 12B 15Cohere Command R+ (Aug. 2024) 25Claude 3 Haiku
87.5%	Non-passive narration preserved	100Claude Sonnet 4.6 100Claude Sonnet 4 100Claude Haiku 4.5	0Claude 3 Haiku 9Rocinante 12B 11Cohere Command R+ (Aug. 2024)
74.5%	Passive → active voice transformations	98GPT-5 98Claude Sonnet 4.5 97GPT-5.1	0Claude 3 Haiku 0Gemma 3 12B 1Rocinante 12B
100.0%	Structural similarity to original	100Claude Sonnet 4.6 (Reasoning) 100GPT-5 Mini 100Gemini 2.5 Flash (Reasoning)	67Claude 3 Haiku 72WizardLM 2 8x22b 78Mistral NeMO

Text Replacement

Passive voice → active voice

Performance Score Distribution (Top 20)

Price-Performance Score Distribution (Top 20)

Most Stable Models (Top 20)

Top Overall Models (Top 20)