Liwati menyang isi utama
OpenAI

27 Februari 2025

RilisProduk

Introducing GPT‑4.5

A research preview of our strongest GPT model. Available to Pro users and developers worldwide.

Coba ing ChatGPT(mbukak ing jendhela anyar)
Lagi dimuat…

Kita lagi ngrilis pratayang riset GPT‑4.5—model paling gedhe lan paling apik kanggo chat nganti saiki. GPT‑4.5 minangka langkah maju ing nggedhekake latihan awalan lan pascalatihan. Kanthi nggedhekake sinau tanpa pengawasan, GPT‑4.5 nambah kemampuan kanggo ngenali pola, narik sambungan, lan ngasilake wawasan kreatif tanpa nalar.

Pengujian awal nuduhake yen sesambungan karo GPT‑4.5 krasa luwih alami. Basis kawruh sing luwih jembar, kemampuan sing luwih apik kanggo nuruti maksud pangguna, lan “EQ” sing luwih gedhe nggawe migunani kanggo tugas kaya ningkatake tulisan, pemrograman, lan ngrampungake masalah praktis. Kita uga ngarepake model iki bakal luwih sithik ngalami halusinasi.

Kita nuduhake GPT‑4.5 minangka pratayang riset kanggo luwih mangerteni kekuwatan lan watesane. Kita isih njelajah apa wae sing bisa ditindakake lan kepengin weruh carane wong nggunakake kanthi cara sing bisa wae ora kita kira sadurunge.

Nggedhekake sinau tanpa pengawasan

Kita ngembangake kemampuan AI kanthi nggedhekake rong paradigma sing saling nglengkapi: sinau tanpa pengawasan lan nalar. Loro iki makili rong sumbu kecerdasan.

  1. Sinau tanpa pengawasan nambah akurasi model donya lan intuisi. Model kaya GPT‑3.5, GPT‑4, lan GPT‑4.5 ngembangake paradigma iki.
  2. Nggedhekake nalar⁠, ing sisih liya, mulang model supaya mikir lan ngasilake rantai pikiran sadurunge mangsuli, saengga bisa ngrampungake masalah STEM utawa logika sing rumit. Model kaya OpenAI o1 lan OpenAI o3‑mini ngembangake paradigma iki.

GPT‑4.5 iku conto nggedhekake sinau tanpa pengawasan kanthi nambah komputasi lan data, bebarengan karo inovasi arsitektur lan optimisasi. GPT‑4.5 dilatih ing superkomputer Microsoft Azure AI. Asile yaiku model sing nduweni kawruh luwih amba lan pangerten luwih jero marang donya, sing nyebabake halusinasi luwih sithik lan reliabilitas luwih apik ing macem-macem topik.

Nggedhekake paradigma GPT

What was the first language?

We don't know exactly what the first language was, and scientists believe we'll probably never know for sure because spoken languages appeared long before written records existed.

Here's what we do know:

  • No single "first language": Languages likely developed gradually in different groups of humans around the same time, tens or even hundreds of thousands of years ago.
  • Origin in Africa: Most scientists agree that spoken language probably first emerged among humans in Africa, perhaps over 100,000 years ago.
  • Proto-languages: Rather than a single language, humans likely started with simple communication that evolved into more complex languages. The earliest forms of language probably included gestures, sounds, and signals to express basic ideas and emotions.
  • Lost to history: Because speech leaves no direct fossil record, and writing emerged relatively recently (around 5,000 years ago), we have no way to know exactly how the earliest languages sounded.

Today, linguists study existing languages to understand how they evolved over thousands of years, but the exact identity of humanity’s very first language remains—and will likely always remain—a mystery.

Kawruh donya luwih jero

Akurasi SimpleQA (luwih gedhe luwih apik)
62.5%38.2%47%15%Akurasi
Tingkat Halusinasi SimpleQA (luwih cilik luwih apik)
37.1%61.8%44%80.3%Tingkat halusinasi

SimpleQA ngukur kefaktualan LLM (large language model) ing pitakon kawruh sing langsung nanging nantang.

Latihan kanggo kolaborasi manungsa

Nalika kita nggedhekake model lan model bisa ngrampungake masalah sing luwih rumit, dadi saya penting kanggo mulangake pangerten sing luwih gedhe babagan kabutuhan lan maksud manungsa. Kanggo GPT‑4.5, kita ngembangake teknik anyar sing bisa diskalakake kanggo nglatih model sing luwih gedhe lan luwih kuat nganggo data sing asalé saka model luwih cilik. Teknik iki ningkatake steerability GPT‑4.5, pangerten marang nuansa, lan obrolan alami.

Evaluasi komparatif karo tester manungsa
56.8%63.2%57.0%Tingkat menang GPT-4.5 vs GPT-4o

Preferensi manungsa ngukur persentase pitakon nalika para tester milih GPT‑4.5 tinimbang GPT‑4o.

Nggabungake pangerten jero babagan donya karo kolaborasi sing luwih apik ngasilake model sing nggabungake gagasan kanthi alami ing obrolan sing anget lan intuïtif, luwih selaras karo kolaborasi manungsa. GPT‑4.5 nduweni pangerten luwih apik babagan apa sing dimaksud manungsa lan napsirake isyarat alus utawa pangarepan tersirat kanthi luwih nyandra lan “EQ” luwih dhuwur. GPT‑4.5 uga nuduhake intuisi estetis lan kreativitas sing luwih kuwat. Model iki unggul kanggo mbantu nulis lan desain.

Kasus panggunaan

GPT-4.5

I’m going through a tough time after failing a test

Aw, I'm really sorry to hear that. Failing a test can feel pretty tough and discouraging, but remember, it's just one moment—not a reflection of who you are or your capabilities. Want to talk about what happened, or do you just need a distraction? I'm here either way.

I’m going through a tough time after failing a test

I'm really sorry to hear that you're going through a tough time. Here are a few things you might consider:

1. Reflect and Understand: Try to analyze what might have gone wrong. Was it a lack of preparation, time management during the test, or perhaps misunderstanding the material? Understanding the root cause can help you address it in the future.

2. Seek Support: Talk to someone you trust—friends, family, or a school counselor. They can offer a different perspective and sometimes just talking about your feelings can be alleviating.

3. Speak with Your Teacher or Professor: They can provide insights into where you might have gone wrong and offer advice on how to improve. They might also offer opportunities for make-up work or extra credit.

4. Develop a Study Plan: Based on your reflections, create a plan to improve your study habits. This might include setting regular study times, breaking down material into manageable portions, or finding new study methods that work better for you.

5. Stay Positive and Forgive Yourself: Mistakes and setbacks are a part of life. It’s important to be kind to yourself and recognize that failure is a stepping stone to success.

6. Focus on the Bigger Picture: Remember that one test is just one part of your educational journey. There will be many more opportunities to do well.

If you need further support or resources, consider reaching out to educational support services at your institution, or mental health resources if you’re feeling particularly overwhelmed. You’re not alone in this, and things can get better with time and effort.

GPT‑4.5 nuduhake “EQ” sing luwih gedhe lan ngerti kapan kudu ngajak obrolan luwih lanjut lan kapan kudu menehi pangguna informasi sing jembar.

Nalar luwih kuwat ing cakrawala

GPT‑4.5 ora mikir dhisik sadurunge nanggapi, mula kekuwatane beda banget saka model nalar kaya OpenAI o1. Dibandhingake OpenAI o1 lan OpenAI o3‑mini, GPT‑4.5 iku model sing luwih umum panggunaane lan luwih pinter kanthi alami. Kita percaya nalar bakal dadi kemampuan inti model masa depan, lan loro pendekatan nggedhekake—latihan awalan lan nalar—bakal saling nglengkapi. Nalika model kaya GPT‑4.5 dadi luwih pinter lan luwih ngerti liwat latihan awalan, model iki bakal dadi dhasar sing luwih kuwat kanggo agen sing nggunakake nalar lan piranti.

Keamanan

Saben peningkatan kemampuan model uga dadi kesempatan kanggo nggawe model luwih aman. GPT‑4.5 dilatih nganggo teknik pengawasan anyar sing digabung karo panyetel diawasi tradisional (SFT) lan metode Sinau Penguatan saka Umpan Balik Manungsa (RLHF) kaya sing digunakake kanggo GPT‑4o. Muga-muga karya iki dadi dhasar kanggo nyelarasake model masa depan sing luwih mumpuni.

Kanggo nguji kanthi ketat perbaikan iki, kita nindakake rangkaian tes keamanan sadurunge peluncuran, selaras karo Kerangka Kesiapan⁠(mbukak ing jendhela anyar)⁠. Kita nemokake yen nggedhekake paradigma GPT nyumbang marang peningkatan kemampuan ing saindenging evaluasi kita. Kita nerbitake asil rinci saka evaluasi iki ing kertu sistem⁠ sing ndherek.

Cara nggunakake GPT-4.5 ing ChatGPT

Wiwit dina iki, pangguna ChatGPT Pro bakal bisa milih GPT‑4.5 ing pemilih model ing web, mobile, lan desktop. Kita bakal miwiti peluncuran kanggo pangguna Plus lan Team minggu ngarep, banjur kanggo pangguna Enterprise lan Edu ing minggu sabanjure.

GPT‑4.5 nduweni akses menyang informasi paling anyar liwat search, ndhukung unggahan file lan gambar, lan bisa nggunakake canvas kanggo nggarap tulisan lan kode. Nanging, GPT‑4.5 saiki durung ndhukung fitur multimodal kaya mode swara, video, lan nuduhake layar ing ChatGPT. Ing mangsa ngarep, kita bakal ngupaya nyederhanakake pengalaman pangguna supaya AI “mung bisa mlaku” kanggo sampeyan.

Cara nggunakake GPT-4.5 ing API

Kita uga lagi nampilake pratayang GPT‑4.5 ing chat completions API, Assistants API, lan Batch API kanggo para pangembang ing kabeh tingkat panggunaan mbayar⁠(mbukak ing jendhela anyar). Model iki ndhukung fitur utama kaya nelpon fungsi, keluaran terstruktur, streaming, lan pesen sistem. Uga ndhukung kemampuan visi liwat input gambar.

Adhedhasar pengujian awal, para pangembang bisa nganggep GPT‑4.5 migunani banget kanggo aplikasi sing entuk manfaat saka kecerdasan emosional lan kreativitas sing luwih dhuwur—kayata pitulungan nulis, komunikasi, sinau, coaching, lan gagasan bareng. Model iki uga nuduhake kemampuan kuwat ing perencanaan lan eksekusi agentic, kalebu alur kerja coding multi-langkah lan otomatisasi tugas rumit.

GPT‑4.5 iku model gedhe banget lan mbutuhake komputasi intensif, mula luwih larang⁠ tinimbang lan dudu pengganti kanggo GPT‑4o. Mula saka iku, kita lagi ngevaluasi apa bakal terus nyedhiyakake ing API kanggo jangka panjang nalika nyimbangake dhukungan kanggo kemampuan saiki karo mbangun model masa depan. Kita ngarep bisa sinau luwih akeh babagan kekuwatan, kemampuan, lan potensi aplikasine ing kahanan nyata. Yen GPT‑4.5 menehi nilai unik kanggo kasus panggunaan sampeyan, umpan balik⁠(mbukak ing jendhela anyar) sampeyan bakal nduweni peran penting kanggo nuntun keputusan kita.

Panutup

Saben ana kenaikan urutan besaran komputasi, muncul kemampuan anyar. GPT‑4.5 iku model ing garis tercanggih saka apa sing bisa digayuh ing sinau tanpa pengawasan. Kita terus kaget karo kreativitas komunitas nalika nemokake kemampuan anyar lan kasus panggunaan sing ora dikira. Kanthi GPT‑4.5, kita ngajak sampeyan njelajah garis tercanggih sinau tanpa pengawasan lan nemokake kemampuan anyar bebarengan karo kita.

Lampiran

Ing ngisor iki, kita nyedhiyakake asil GPT‑4.5 ing benchmark akademik standar kanggo nggambarake performa saiki ing tugas sing biasane digandhengake karo nalar. Sanajan mung kanthi nggedhekake sinau tanpa pengawasan, GPT‑4.5 nuduhake peningkatan sing teges dibandhing model sadurunge kaya GPT‑4o. Nanging, kita ngarep bisa entuk gambaran luwih lengkap babagan kemampuan GPT‑4.5 liwat rilis iki, amarga kita ngerti benchmark akademik ora mesthi nggambarake migunani ing donya nyata.

Skor evaluasi model

GPT‑4.5

GPT‑4o

OpenAI o3‑mini (dhuwur)

GPQA (sains)

71.4%

53.6%

79.7%

AIME ‘24 (matematika)

36.7%

9.3%

87.3%

MMMLU (multibasa)

85.1%

81.5%

81.1%

MMMU (multimodal)

74.4%

69.1%

-

SWE-Lancer Diamond (coding)*

32.6%

$186,125

23.3%

$138,750

10.8%

$89,625

SWE-Bench Verified (coding)*

38.0%

30.7%

61.0%

*Angka sing ditampilake makili performa internal paling apik.

Panulis

OpenAI

Kontributor dhasar

Adam Goucher, Alex Paino, Ali Kamali, Amin Tootoonchian, Andrew Tulloch, Ben Sokolowsky, Clemens Winter, Colin Wei, Daniel Kappler, Daniel Levy, Felipe Petroski Such, Geoff Salmon, Ian O’Connell, Jason Teplitz, Kai Chen, Nik Tezak, Prafulla Dhariwal, Rapha Gontijo Lopes, Sam Schoenholz, Youlong Cheng, Yujia Jin, Yunxing Dai

Riset

Kontributor inti

Aiden Low, Alec Radford, Alex Carney, Alex Nichol, Alexis Conneau, Ananya Kumar, Ben Wang, Charlotte Cole , Elizabeth Yang, Gabriel Goh, Hadi Salman, Haitang Hu, Heewoo Jun, Ian Sohl, Ishaan Gulrajani, Jacob Coxon, James Betker, Jamie Kiros, Jessica Landon, Kyle Luther, Lia Guy, Lukas Kondraciuk, Lyric Doshi, Mikhail Pavlov, Qiming Yuan, Reimar Leike, Rowan Zellers, Sean Metzger, Shengjia Zhao, Spencer Papay, Tao Wang

Kontributor

Adam Lerer, Adrien Ecoffet, Aidan McLaughlin, Alexander Prokofiev, Alexandra Barr, Allan Jabri, Andrew Gibiansky, Andrew Schmidt, Casey Chu, Chak Li, Chelsea Voss, Chris Hallacy, Chris Koch, Christine McLeavey, David Mely, Dimitris Tsipras, Eric Sigler, Erin Kavanaugh, Farzad Khorasani, Huiwen Chang, Ilya Kostrikov, Ishaan Singal, Ji Lin, Jiahui Yu, Jing Yu Zhang, John Rizzo, Jong Wook Kim, Joyce Lee, Juntang Zhuang, Leo Liu, Li Jing, Long Ouyang, Louis Feuvrier, Mo Bavarian, Nick Stathas, Nitish Keskar, Oleg Murk, Preston Bowman, Scottie Yan, SQ Mah, Tao Xu, Taylor Gordon, Valerie Qi, Wenda Zhou, Yu Zhang

Nggedhekake

Kontributor inti

Alex Chow, Alex Renzin, Aleksandra Spyra, Avi Nayak, Ben Leimberger, Christopher Hesse, Duc Phong Nguyen, Dinghua Li, Eric Peterson, Francis Zhang, Gene Oden, Kai Fricke, Kai Hayashi, Larry Lv, Leqi Zou, Lin Yang, Madeleine Thompson, Michael Petrov, Miguel Castro, Natalia Gimelshein, Phil Tillet, Reza Zamani, Ryan Cheu Stanley Hsieh, Steve Lee, Stewart Hall, Thomas Raoux, Tianhao Zheng, Vishal Kuo, Yongjik Kim, Yuchen Zhang, Zhuoran Liu

Kontributor

Alvin Wan, Andrew Cann, Andrew Codispoti, Antoine Pelisse, Anuj Kalia, Aaron Hurst, Avital Oliver, Brad Barnes, Brian Hsu, Chen Ding, Chen Shen, Cheng Chang, Christian Gibson, Christopher Berner, Duncan Findlay, Fan Wang, Fangyuan Li, Gianluca Borello, Heather Schmidt, Henrique Ponde de Oliveira Pinto, Ikai Lan, Jiayi Weng, James Crooks, Jos Kraaijeveld, Junru Shao, Kenny Hsu, Kenny Nguyen, Kevin King, Leah Burkhardt, Leo Chen, Linden Li, Lu Zhang, Mahmoud Eariby, Marat Dukhan, Mateusz Litwin, Miki Habryn, Natan LaFontaine, Pavel Belov, Peng Su, Prasad Chakka, Rachel Lim, Rajkumar Samuel, Renaud Gaubert, Rory Carmichael, Sarah Dong, Shantanu Jain, Shuaiqi Xia, Stephen Logsdon, Todd Underwood, Tony Zhao, Weixing Zhang, Will Sheu, Weiyi Zheng, Yinghai Lu, Yunqiao Zhang

Sistem Keamanan

Andrea Vallone, Andy Applebaum, Cameron Raymond, Chong Zhang, Dan Mossing, Elizabeth Proehl, Eric Wallace, Evan Mays, Grace Zhao, Ian Kivlichan, Irina Kofman, Joel Parish, Kevin Liu, Keren Gu-Lemberg, Kristen Ying, Lama Ahmad, Lilian Weng, Leon Maksin, Leyton Ho, Meghan Shah, Michael Lampe, Michele Wang, Miles Wang, Olivia Watkins, Phillip Guo, Samuel Miserendino, Sam Toizer, Sandhini Agarwal, Tejal Patwardhan, Tom Dupré la Tour, Tong Mu, Tyna Eloundou, Yunyun Wang

Peluncuran

Adam Brandon, Adam Perelman, Adele Li, Akshay Nathan, Alan Hayes, Alfred Xue, Alison Ben, Alec Gorge, Alex Guziel, Alex Iftimie, Ally Bennett, Andrew Chen, Andy Wang, Andy Wood, Angad Singh, Anoop Kotha, Antonia Woodford, Anuj Saharan, Ashley Tyra, Atty Eleti, Ben Schneider, Bessie Ji, Beth Hoover, Bill Chen, Blake Samic, Britney Smith, Brian Yu, Caleb Wang, Cary Bassin, Cary Hudson, Charlie Jatt, Chengdu Huang, Chris Beaumont, Christina Huang, Cristina Scheau, Dana Palmie, Daniel Levine, Daryl Neubieser, Dave Cummings, David Sasaki, Dibya Bhattacharjee, Dylan Hunn, Edwin Arbus, Elaine Ya Le, Enis Sert, Eric Kramer, Fred von Lohmann, Freddie Sulit, Gaby Janatpour, Garrett McGrath, Garrett Ollinger, Gary Yang, Hao Sheng, Harold Hotelling, Janardhanan Vembunarayanan, Jeff Harris, Jeffrey Sabin Matsumoto, Jennifer Robinson, Jessica Liang, Jessica Shieh, Jiacheng Yang, Joel Morris, Joseph Florencio, Josh Kaplan, Kan Wu, Karan Sharma, Karen Li, Katie Pypes, Kendal Simon, Kendra Rimbach, Kevin Park, Kevin Rao, Laurance Fauconnet, Lauren Workman, Leher Pathak, Liang Wu, Liang Xiong, Lien Mamitsuka, Lindsay McCallum, Lukas Gross, Manoli Liodakis, Matt Nichols, Michelle Fradin, Minal Khan, Mingxuan Wang, Nacho Soto, Natalie Staudacher, Nikunj Handa, Niko Felix, Ning Liu, Olivier Godement, Oona Gleeson, Philip Pronin, Raymond Li, Reah Miyara, Robert Xiong, Rohan Nuttall, R.J. Marsan, Sara Culver, Scott Ethersmith, Sean Fitzgerald, Shamez Hemani, Sherwin Wu, Shiao Lee, Shuyang Cheng, Siyuan Fu, Spug Golden, Steve Coffey, Steven Heidel, Sundeep Tirumalareddy, Tabarak Khan, Thomas Degry, Thomas Dimson, Tom Stasi, Tomo Hiratsuka, Trevor Creech, Uzair Navid Iftikhar, Victoria Chernova, Victoria Spiegel, Wanning Jiang, Wenlei Xie, Yaming Lin, Yara Khakbaz, Yilei Qian, Yilong Qin, Yo Shavit, Zhi Bie

Pimpinan Eksekutif

Aidan Clark, Bob McGrew, David Farhi, Greg Brockman, Hannah Wong, Jakub Pachocki, Johannes Heidecke, Joanne Jang, Kate Rouch, Kevin Weil, Lauren Itow, Liam Fedus, Mark Chen, Mia Glaese, Mira Murati, Nick Ryder, Sam Altman, Srinivas Narayanan, Tal Broda