参与模型

共 24 个参与模型（24 个测试记录）。

排名	模型	测试次数	平均运行成本	平均输出量	平均执行步数	最高能力得分
1	claude-opus-5 max	1	$11.84	118k	99	73.65%
2	gpt-5-6-sol max	1	$8.39	60k	61	72.67%
3	claude-fable-5 xhigh	1	$13.41	80k	68	69.91%
4	gpt-5-6-terra max	1	$4.95	72k	76	69.62%
5	kimi-k3 max	1	$4.65	81k	98	68.51%
6	grok-4-6 medium	1	$3.45	50k	70	67.48%
7	gpt-5-6-luna max	1	$3.03	73k	102	67.19%
8	gpt-5-5 xhigh	1	$7.23	46k	82	67.04%
9	gemini-3-7-flash medium	1	$2.03	94k	117	65.49%
10	deepseek-v4-pro max	1	$0.24	106k	155	62.83%
11	claude-opus-4-8 max	1	$13.22	135k	120	58.97%
12	qwen3-8-max xhigh	1	$3.73	95k	111	57.46%
13	muse-spark-1-2 xhigh	1	$3.70	99k	101	54.87%
14	claude-sonnet-5 max	1	$26.40	214k	268	53.85%
15	grok-4-5 high	1	$2.42	36k	61	53.76%
16	deepseek-v4-flash max	1	$0.10	108k	153	53.32%
17	muse-spark-1-1 xhigh	1	$2.36	74k	96	53.32%
18	gpt-5-4 xhigh	1	$5.65	71k	70	51.77%
19	gemini-3-6-flash high	1	$4.42	96k	117	46.68%
20	glm-5-2 max	1	$3.92	78k	129	43.78%
21	gemini-3-5-flash high	1	$3.45	76k	105	36.06%
22	kimi-k2-7-code	1	$2.82	59k	149	30.53%
23	claude-sonnet-4-6 high	1	$5.52	76k	134	29.93%
24	gemini-3-1-pro-preview high	1	$2.14	28k	76	11.73%