Aider Polyglot 代码编辑榜

名次与分数均取自官方榜单，与官方一致。条形长度按当前这些成绩里的最大值算，领跑那条填满整列，其余按比例缩短（灰斜纹是「额外真实缺陷」，只与同列比、不与分数比）。官方没给的数据显示「—」。

Aider Polyglot 代码编辑榜 通过率排名

名次	模型	一次通过	评测成本（$）	单例耗时（s）	通过率	说明
1	gpt-5 (high)	52	29.1	194	88	diff · high · 2025-08-23
2	gpt-5 (medium)	49.8	17.7	118.7	86.7	diff · medium · 2025-08-25
3	o3-pro (high)	43.6	146.3	449	84.9	diff · high · 2025-06-28
4	gemini-2.5-pro-preview-06-05 (32k think)	46.2	49.9	200.3	83.1	diff-fenced · 2025-06-06
5	o3 (high)	40	21.2	197.3	81.3	diff · high · 2025-06-25
6	gpt-5 (low)	43.1	10.4	62.4	81.3	diff · low · 2025-08-25
7	grok-4 (high)	40.9	59.6	403.2	79.6	diff · high · 2025-07-11
8	gemini-2.5-pro-preview-06-05 (default think)	44.9	45.6	175.2	79.1	diff-fenced · 2025-06-06
9	o3 (high) + gpt-4.1	34.8	17.6	121.8	78.2	architect · high · 2025-06-27
10	Gemini 2.5 Pro Preview 05-06	36.4	37.4	165.3	76.9	diff-fenced · 2025-05-07
11	o3	40.9	13.8	101.7	76.9	diff · 2025-06-25
12	DeepSeek-V3.2-Exp (Reasoner)	39.6	1.3	291.2	74.2	diff · 2025-10-03
13	Gemini 2.5 Pro Preview 03-25	40.9	0	45.3	72.9	diff-fenced · 2025-04-12
14	o4-mini (high)	19.6	19.6	176.5	72	diff · 2025-04-16
15	claude-opus-4-20250514 (32k thinking)	37.3	65.7	44.1	72	diff · 2025-05-25
16	DeepSeek R1 (0528)	34.4	4.8	716.6	71.4	diff · 2025-06-06
17	claude-opus-4-20250514 (no think)	32.9	68.6	42.5	70.7	diff · 2025-05-25
18	DeepSeek-V3.2-Exp (Chat)	38.7	0.9	104	70.2	diff · 2025-10-03
19	claude-3-7-sonnet-20250219 (32k thinking tokens)	29.3	36.8	105.2	64.9	diff · 2025-02-24
20	DeepSeek R1 + claude-3-5-sonnet-20241022	27.1	13.3	251.6	64	architect · 2025-01-23
21	o1-2024-12-17 (high)	23.7	186.5	133.2	61.7	diff · 2024-12-21
22	claude-sonnet-4-20250514 (32k thinking)	25.8	26.6	79.9	61.3	diff · 2025-05-24
23	o3-mini (high)	21	18.2	124.6	60.4	diff · 2025-01-31
24	claude-3-7-sonnet-20250219 (no thinking)	24.4	17.7	28.3	60.4	diff · 2025-02-24
25	Qwen3 235B A22B diff, no think, Alibaba API	28.9	0	45.4	59.6	diff · 2025-05-09
26	Kimi K2	20.4	1.2	67.6	59.1	diff · 2025-07-17
27	DeepSeek R1	26.7	5.4	113.7	56.9	diff · 2025-01-20
28	claude-sonnet-4-20250514 (no thinking)	20.4	15.8	29.8	56.4	diff · 2025-05-24
29	DeepSeek V3 (0324)	28	1.1	290	55.1	diff · 2025-03-24
30	gemini-2.5-flash-preview-05-20 (24k think)	26.2	8.6	53.9	55.1	diff · 2025-05-25
31	Quasar Alpha	21.8	0	14.8	54.7	diff · 2025-04-04
32	o3-mini (medium)	19.1	8.9	47.2	53.8	diff · 2025-01-31
33	Grok 3 Beta	22.2	11	15.3	53.3	diff · 2025-04-10
34	Optimus Alpha	21.3	0	18.4	52.9	diff · 2025-04-10
35	gpt-4.1	20	9.9	20.5	52.4	diff · 2025-04-14
36	claude-3-5-sonnet-20241022	22.2	14.4	21.4	51.6	diff · 2025-01-17
37	Grok 3 Mini Beta (high)	17.3	0.7	79.1	49.3	whole · 2025-04-10
38	DeepSeek Chat V3 (prev)	22.7	0.3	34.8	48.4	diff · 2024-12-25
39	gemini-2.5-flash-preview-04-17 (default)	21.8	1.8	50.1	47.1	diff · 2025-04-20
40	chatgpt-4o-latest (2025-03-29)	16.4	19.7	10.3	45.3	diff · 2025-03-29
41	gpt-4.5-preview	22.3	183.2	113.5	44.9	diff · 2025-02-27
42	gemini-2.5-flash-preview-05-20 (no think)	20.9	1.1	12.2	44	diff · 2025-05-26
43	gpt-oss-120b (high)	13.8	0.7	35.5	41.8	diff · 2025-08-06
44	Qwen3 32B	14.2	0.8	372.2	40	diff · 2025-05-08
45	gemini-exp-1206	19.6	0	45.5	38.2	whole · 2024-12-22
46	Gemini 2.0 Pro exp-02-05	20.4	0	34.8	35.6	whole · 2025-02-25
47	Grok 3 Mini Beta (low)	11.1	0.8	35.1	34.7	whole · 2025-04-10
48	o1-mini-2024-09-12	5.8	18.6	34.7	32.9	whole · 2024-12-22
49	gpt-4.1-mini	11.1	2	19.5	32.4	diff · 2025-04-14
50	claude-3-5-haiku-20241022	7.1	6.1	31.8	28	diff · 2024-12-21
51	chatgpt-4o-latest (2025-02-15)	9	14.4	12.4	27.1	diff · 2025-02-15
52	QwQ-32B + Qwen 2.5 Coder Instruct	9.8	0	137.4	26.2	architect · 2025-03-07
53	gpt-4o-2024-08-06	4.9	7	16	23.1	diff · 2024-12-30
54	gemini-2.0-flash-exp	11.6	0	12.2	22.2	whole · 2024-12-22
55	qwen-max-2025-01-25	9.3	—	39.5	21.8	diff · 2025-01-28
56	QwQ-32B	8	0	228.6	20.9	diff · 2025-03-06
57	gpt-4o-2024-11-20	4.9	6.7	12.1	18.2	diff · 2024-12-30
58	gemini-2.0-flash-thinking-exp-01-21	5.8	0	24.2	18.2	diff · 2025-01-21
59	DeepSeek Chat V2.5	5.3	0.5	184	17.8	diff · 2024-12-21
60	Qwen2.5-Coder-32B-Instruct	4.9	0	42	16.4	whole · 2024-12-26
61	Llama 4 Maverick	4.4	0	20.5	15.6	whole · 2025-04-06
62	yi-lightning	5.8	0	146.7	12.9	whole · 2024-12-23
63	command-a-03-2025-quality	2.2	0	85.1	12	whole · 2025-03-14
64	Codestral 25.01	4	2	9.3	11.1	whole · 2025-01-13
65	openhands-lm-32b-v0.1	4	0	195.6	10.2	whole · 2025-04-19
66	gpt-4.1-nano	3.1	0.4	12	8.9	whole · 2025-04-14
67	gemma-3-27b-it	1.8	0	79.7	4.9	whole · 2025-03-15
68	gpt-4o-mini-2024-07-18	0.9	0.3	17.3	3.6	whole · 2024-12-21

展示分用的是 pass_rate_2（给一次重试机会），一次通过率另列。 同一模型多次提交只保留通过率最高的一次，附加指标说明里标注了编辑格式与日期。