OpenAI (AI company)OpenAIOpenAIto releasefābù发布GPT-5.4 (OpenAI language model)GPT-5.4GPT-5.4，to supportzhīchí支持one millionbǎiwàn百万tokenscí yuán词元context windowshàngxià wén上下文andbìng并cankě可to operate / controlcāokòng操控computerdiànnǎo电脑。
OpenAI (AI company)
OpenAI
OpenAI
to release
fābù
发布
GPT-5.4 (OpenAI language model)
GPT-5.4
GPT-5.4
to support
zhīchí
支持
one million
bǎiwàn
百万
tokens
cí yuán
词元
context window
shàngxià wén
上下文
and
bìng
并
can
kě
可
to operate / control
cāokòng
操控
computer
diànnǎo
电脑
OpenAI (AI company)
OpenAI
OpenAI
on (formal)
yú
于
March 5, 2026
2026 nián 3 yuè 5 rì
2026年3月5日
released
fābù
发布
completion marker
le
了
latest
zuìxīn
最新
flagship
qíjiàn
旗舰
model
móxíng
模型
GPT-5.4 (OpenAI language model)
GPT-5.4
GPT-5.4
the new model
xīn móxíng
新模型
supports
zhīchí
支持
up to
gāodá
高达
one million
yī bǎi wàn gè
一百万个
tokens
cí yuán
词元
structural particle
de
的
context window
shàngxià wén chuāngkǒu
上下文窗口
and
bìng
并
can
kě
可
autonomously
zìzhǔ
自主
to operate
cāokòng
操控
computer
diànnǎo
电脑
and
hé
和
web browser
liúlǎn qì
浏览器
March 5, 2026 at 8:00 PM
OpenAI (AI company)
OpenAI
OpenAI
on (formal)
yú
于
March 5, 2026
2026 nián 3 yuè 5 rì
2026年3月5日
officially
zhèngshì
正式
launched
tuīchū
推出
completion marker
le
了
its
qí
其
latest
zuìxīn
最新
flagship
qíjiàn
旗舰
model
móxíng
模型
GPT-5.4 (OpenAI language model)
GPT-5.4
GPT-5.4
the (formal)
gāi
该
model
móxíng
模型
passive marker
bèi
被
is positioned as
dìngwèi wéi
定位为
currently
mùqián
目前
the most powerful
zuì qiángdà
最强大
the most efficient
zuì gāoxiào
最高效
structural particle
de
的
professional work
zhuānyè gōngzuò
专业工作
frontier model
qiányán móxíng
前沿模型
in
zài
在
API (application programming interface)
API
API
version
bǎnběn
版本
in
zhōng
中
GPT-5.4 (OpenAI language model)
GPT-5.4
GPT-5.4
supports
zhīchí
支持
up to
gāodá
高达
one million
yī bǎi wàn gè
一百万个
tokens
cí yuán
词元
structural particle
de
的
context window
shàngxià wén chuāngkǒu
上下文窗口
setting
chuàng
创
a historic new high
lìshǐ xīn gāo
历史新高
compared with
yǔ
与
the previous generation model
qián dài móxíng
前代模型
compared
xiāngbǐ
相比
the new model
xīn móxíng
新模型
in
zài
在
individual claims
dān tiáo xìnxī
单条信息
error rate
cuòwù lǜ
错误率
in terms of
shàng
上
reduced by
jiàngdī le
降低了
33 percent
33%
33%
overall
zhěngtǐ
整体
responses
huídá
回答
containing errors
chūcuò
出错
structural particle
de
的
probability
gàilǜ
概率
also
yě
也
decreased by
jiǎnshǎo le
减少了
18 percent
18%
18%
in addition
cǐwài
此外
GPT-5.4 (OpenAI language model)
GPT-5.4
GPT-5.4
integrated
zhěnghé le
整合了
code generation
dàimǎ shēngchéng
代码生成
advanced reasoning
gāojí tuīlǐ
高级推理
and
yǔ
与
desktop control
zhuōmiàn cāokòng
桌面操控
and other
děng
等
multiple
duō xiàng
多项
capabilities
nénglì
能力
the model
gāi móxíng
该模型
can
kěyǐ
可以
autonomously
zìzhǔ
自主
browse web pages
liúlǎn wǎngyè
浏览网页
operate
cāozuò
操作
desktop applications
zhuōmiàn chéngxù
桌面程序
this capability
zhè yī gōngnéng
这一功能
is called
bèi chēngwéi
被称为
computer use
jìsuànjī shǐyòng
计算机使用
ability
nénglì
能力
OpenAI (AI company)
OpenAI
OpenAI
at the same time
tóngshí
同时
also released
tuīchū le
推出了
smaller
gèng xiǎo de
更小的
GPT-5.4 Mini (smaller model)
GPT-5.4 Mini
GPT-5.4 Mini
and
hé
和
Nano (smallest model tier)
Nano
Nano
versions
bǎnběn
版本
to meet
yǐ mǎnzú
以满足
different
bùtóng
不同
users
yònghù
用户
structural particle
de
的
needs
xūqiú
需求
analysts
fēnxī rénshì
分析人士
point out
zhǐchū
指出
this time
cǐ cì
此次
release
fābù
发布
further
jìnyībù
进一步
expanded
kuòdà le
扩大了
OpenAI (AI company)
OpenAI
OpenAI
in
zài
在
AI (artificial intelligence)
AI
AI
field
lǐngyù
领域
structural particle
de
的
competitive advantage
jìngzhēng yōushì
竞争优势
New Words
tuīchū
推出
to release / launch
móxíng
模型
model
zuìxīn
最新
latest
qíshuō
旗舰
flagship
qiángdà
强大
powerful
gāoxiào
高效
efficient
zhuānyè
专业
professional
gōngzuò
工作
work
tuīlǐ
推理
reasoning
dàimǎ
代码
code
News in English
Science & TechnologyOpenAI Releases GPT-5.4 With One-Million-Token Context Window and Native Computer Use
OpenAI released GPT-5.4 on March 5, 2026, its most capable frontier model to date. The model supports context windows of up to one million tokens and can autonomously operate computers and web browsers.
OpenAI officially launched its latest flagship model GPT-5.4 on March 5, 2026. The model is positioned as the most powerful and efficient frontier model for professional work. In the API version, GPT-5.4 supports high-quality, long-context processing with unprecedented capabilities in reasoning and code generation.
The release of GPT-5.4 represents a major advancement in large language model technology. The model incorporates improvements in training efficiency, safety, and practical applications across diverse professional fields.
More in Science & Technology
former / previousqián前Groq (AI chip startup)GroqGroqengineersgōngchéngshī工程师regarding / over (a matter)jiù就$20 billion USD200 yì měiyuán200亿美元Nvidia (US semiconductor company)Yīngwěidá英伟达licensing / authorizationshòuquán授权deal / transactionjiāoyì交易to file / to initiatetíqǐ提起a lawsuit / legal actionsùsòng诉讼。
former / previous
qián
前
Groq (AI chip startup)
Groq
Groq
engineers
gōngchéngshī
工程师
regarding / over (a matter)
jiù
就
$20 billion USD
200 yì měiyuán
200亿美元
Nvidia (US semiconductor company)
Yīngwěidá
英伟达
licensing / authorization
shòuquán
授权
deal / transaction
jiāoyì
交易
to file / to initiate
tíqǐ
提起
a lawsuit / legal action
sùsòng
诉讼
Former Groq Engineers Sue Board Over $20 Billion Nvidia Licensing Deal
Qualcomm (US semiconductor company)Gāotōng高通for the first timeshǒucì首次to / towardxiàng向Huawei (Chinese tech company)Huáwéi华为to pay / to make a paymentzhīfù支付patent royalties / licensing feezhuānlì fèi专利费，both partiesshuāngfāng双方to sign / to concludeqiānshǔ签署multi-year periodduō nián qī多年期agreement / dealxiéyì协议。
Qualcomm (US semiconductor company)
Gāotōng
高通
for the first time
shǒucì
首次
to / toward
xiàng
向
Huawei (Chinese tech company)
Huáwéi
华为
to pay / to make a payment
zhīfù
支付
patent royalties / licensing fee
zhuānlì fèi
专利费
both parties
shuāngfāng
双方
to sign / to conclude
qiānshǔ
签署
multi-year period
duō nián qī
多年期
agreement / deal
xiéyì
协议
Qualcomm Pays Huawei for First Time in Broad Multi-Year Patent Licensing Agreement
Tencent (Chinese tech company)Téngxùn腾讯from / towardxiàng向Oracle (US tech corporation)Jiǎgǔwén甲骨文to lease / to rent for usezūyòng租用one hundred thousandshí wàn十万measure word for small round objectskē颗artificial intelligenceréngōng zhìnéng人工智能semiconductor chip / microchipxīnpiàn芯片。
Tencent (Chinese tech company)
Téngxùn
腾讯
from / toward
xiàng
向
Oracle (US tech corporation)
Jiǎgǔwén
甲骨文
to lease / to rent for use
zūyòng
租用
one hundred thousand
shí wàn
十万
measure word for small round objects
kē
颗
artificial intelligence
réngōng zhìnéng
人工智能
semiconductor chip / microchip
xīnpiàn
芯片
Tencent Leases 100,000 AI Chips From Oracle in $7 Billion Deal
Apple Inc.Píngguǒ Gōngsī苹果公司to announcexuānbù宣布will (future marker)jiāng将on / at (preceding a date)yú于October 13thshíyuè shísān rì十月十三日to hold (an event)jǔxíng举行Welcome Home (name of Apple event)Welcome HomeWelcome Homeproduct launch eventfābùhuì发布会，expected toyùjì预计to launch / to introducetuīchū推出HomePad mini (Apple smart home display)HomePad miniHomePad miniand others / etc.děng等new-modelxīnkuǎn新款productschǎnpǐn产品。
Apple Inc.
Píngguǒ Gōngsī
苹果公司
to announce
xuānbù
宣布
will (future marker)
jiāng
将
on / at (preceding a date)
yú
于
October 13th
shíyuè shísān rì
十月十三日
to hold (an event)
jǔxíng
举行
Welcome Home (name of Apple event)
Welcome Home
Welcome Home
product launch event
fābùhuì
发布会
expected to
yùjì
预计
to launch / to introduce
tuīchū
推出
HomePad mini (Apple smart home display)
HomePad mini
HomePad mini
and others / etc.
děng
等
new-model
xīnkuǎn
新款
products
chǎnpǐn
产品
Apple Announces October 13 Welcome Home Event, Expected to Reveal HomePad Mini and Other New Products