Claude Haiku 5.5 vs. gpt-oss-120b：发布比较

Claude Haiku 5.5 与 gpt-oss-120b 发布比较：分别包含 5 个和 2 个模型，各自具有不同的智能、性能与价格特征。下方对全部 7 个模型的关键指标进行了比较。

智能方面，Claude Haiku 5.5 得分最高：Claude Haiku 5.5 (Max)（36），gpt-oss-120b 为 gpt-oss-120b (High)（4）。
每任务耗时方面，Claude Haiku 5.5 最快：Claude Haiku 5.5 (Low)（2.2 分钟），gpt-oss-120b 为 gpt-oss-120b (High)（7.5 分钟）。
单任务成本方面，Claude Haiku 5.5 最低：Claude Haiku 5.5 (Low)（$0.0438），gpt-oss-120b 为 gpt-oss-120b (High)（$0.35）。

比较的发布


	Claude Haiku 5.5 (max)	Claude Haiku 5.5 (xhigh)	Claude Haiku 5.5 (high)	Claude Haiku 5.5 (medium)	Claude Haiku 5.5 (low)	gpt-oss-120b (high)	gpt-oss-120b (low)
Intelligence Index	36	35	30	25	19	4	--
财务与会计	44	41	38	35	29	8	--
战略与运营	39	39	37	33	27	6	--
法律	39	37	35	33	27	8	--
医疗与健康	37	34	32	31	25	6	--
工程	43	41	37	32	27	11	--
经济学	43	41	38	34	29	9	--
AA-Briefcase v1.1	1573	1523	1441	1366	1113	0	--
AutomationBench-AA	35.4%	36.0%	33.7%	28.6%	22.9%	0.2%	--
Terminal-Bench 4.0	32.8%	29.3%	21.7%	15.2%	12.6%	0.0%	--
FrontierCode-AA 2.0	34.9%	29.9%	23.2%	16.1%	8.0%	0.0%	--
AA-Omniscience v1.1	12	8	5	4	2	−46	--
GDP.pdf	20.8%	18.2%	17.2%	15.2%	11.2%	4.2%	--
HLE-Diamond	29.6%	28.8%	23.7%	16.8%	14.1%	3.8%	--
CritPt v2 (Hard Subset)	43.5%	47.5%	45.0%	30.5%	21.5%	4.0%	--
Terminal-Bench-Science 0.1	20.0%	18.1%	10.0%	7.1%	1.9%	0.0%	--

每任务成本美元	$0.41	$0.25	$0.16	$0.0918	$0.0438	$0.35	--
输入价格美元/100 万 Token	$0.10	$0.10	$0.10	$0.10	$0.10	$0.15	$0.13
输出价格美元/100 万 Token	$0.50	$0.50	$0.50	$0.50	$0.50	$0.60	$0.59
缓存命中价格美元/100 万 Token	$0.01	$0.01	$0.01	$0.01	$0.01	$0.13	$0.13
缓存写入价格美元/100 万 Token	$0.13	$0.13	$0.13	$0.13	$0.13	--	--

每任务耗时分钟	14.9	9.8	6.5	4.0	2.2	7.5	--
输出速度tok/s	252	231	224	221	211	164	167
首个答案 Token 延迟秒	174.18	59.27	21.70	10.45	5.87	15.05	2.21
首 Token 延迟秒	174.18	59.27	21.70	10.45	5.87	1.24	1.23
端到端时间单次请求工作负载的平均值（秒）	311.60	118.24	52.55	31.36	23.55	31.78	14.71

上下文窗口	1M	1M	1M	1M	1M	131k	131k
开发商	Anthropic	Anthropic	Anthropic	Anthropic	Anthropic	OpenAI	OpenAI
许可证	专有	专有	专有	专有	专有	开放	开放
输入模态	支持：文本和图像	支持：文本和图像	支持：文本和图像	支持：文本和图像	支持：文本和图像	支持：文本	支持：文本
输出模态	支持：文本	支持：文本	支持：文本	支持：文本	支持：文本	支持：文本	支持：文本
服务商						+12	+11

Updated

智能

Artificial Analysis Intelligence Index · Higher is better

每任务耗时

Inference time (minutes) per Intelligence Index task · Lower is better

每任务成本

Weighted average cost (USD) per Intelligence Index task · Lower is better

智能Updated

Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index v5.0.1 incorporates 9 evaluations: AA-Briefcase v1.1, AutomationBench-AA, Terminal-Bench 4.0, FrontierCode-AA 2.0, AA-Omniscience v1.1, GDP.pdf, HLE-Diamond, CritPt v2 (Hard Subset), and Terminal-Bench-Science 0.1

Artificial Analysis Intelligence Index v5.0.1 includes: AA-Briefcase v1.1, AutomationBench-AA, Terminal-Bench 4.0, FrontierCode-AA 2.0, AA-Omniscience v1.1, GDP.pdf, HLE-Diamond, CritPt v2 (Hard Subset), Terminal-Bench-Science 0.1. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Intelligence Index 与每项任务的成本

Artificial Analysis Intelligence Index · Weighted average cost (USD) per Artificial Analysis Intelligence Index task

Most attractive quadrant

Claude Haiku 5.5

Pareto line

AA-Briefcase v1.1

AutomationBench-AA

单任务成本（美元，对数刻度）

Weighted average cost per Intelligence Index task. Each evaluation’s cost is calculated from input, cache hit, cache write, reasoning, and answer token prices, divided by task count, and weighted by its Intelligence Index weight.

成本

每项 Intelligence Index 任务的成本

Weighted average cost (USD) per Intelligence Index task · Segmented by token type · Lower is better

性能

Intelligence Index 每项任务耗时

Output tokens divided by end-to-end output speed (minutes per Intelligence Index task) · Speed at the model's own endpoint or the median across providers · Lower is better

The weighted average time (minutes) per Artificial Analysis Intelligence Index task: output tokens per task divided by output speed, weighted by each benchmark's weight in the Intelligence Index.

Speed and latency metrics average each endpoint's runs over the past 5 days, and the error rate pools them. We run the benchmark every 6 hours, and endpoints of deprecated models every 18 hours.

能力得分

能力指数

衡量模型在特定能力与行业中的表现

Finance & Accounting Index

Incorporates 5 evaluations: AA-Briefcase v1.1, AA-Omniscience v1.1, HLE-Diamond, AutomationBench-AA, GDP.pdf · Higher is better

Strategy & Ops Index

Incorporates 4 evaluations: AutomationBench-AA, AA-Briefcase v1.1, AA-Omniscience v1.1, GDP.pdf · Higher is better

Legal Index

Incorporates 5 evaluations: AA-Omniscience v1.1, AA-Briefcase v1.1, HLE-Diamond, GDP.pdf, AutomationBench-AA · Higher is better

Healthcare & Medical Index

Incorporates 5 evaluations: AA-Omniscience v1.1, AA-Briefcase v1.1, MLCR-AA, AutomationBench-AA, HLE-Diamond · Higher is better

Engineering Index

Incorporates 7 evaluations: AA-Omniscience v1.1, CritPt v2 (Hard Subset), Terminal-Bench-Science 0.1, HLE-Diamond, Terminal-Bench 4.0, FrontierCode-AA 2.0, AA-Briefcase v1.1 · Higher is better

Economics Index

Incorporates 4 evaluations: AA-Omniscience v1.1, AA-Briefcase v1.1, HLE-Diamond, GDP.pdf · Higher is better

基准测试

AA-PersonalityBenchNew

性格指数

每个轴对应 8 个性格指数之一：其特质 Elo z 分数的平均值，在基准测试中所有已评估模型之间换算为 1 到 100 的百分位排名

Claude Haiku 5.5 vs. gpt-oss-120b：发布比较

智能

每任务耗时

每任务成本

智能Updated

Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index

Intelligence Index 与每项任务的成本

Cost per Intelligence Index Task

Artificial Analysis Intelligence Index

成本

每项 Intelligence Index 任务的成本

Cost per Intelligence Index Task

性能

Intelligence Index 每项任务耗时

Time per Intelligence Index Task

Time Period

能力得分

能力指数

基准测试

AA-PersonalityBenchNew

性格指数