Introducing improvements to the fine-tuning API and expanding our custom models program

Wargelinta 8-da Maayo, 2026: OpenAI waxay si tartiib ah u soo afjaraysaa taageerada madasha u sii tabbabarida AI-ga hawl gaar ah. Madashu hadda uma furna isticmaaleyaal cusub, laakiin isticmaaleyaasha hadda jira ee madasha u sii tabbabarida AI-ga hawl gaar ah waxay awoodi doonaan inay abuuraan shaqooyin tababar bilaha soo socda. Dhammaan moodelada lagu sameeyay ku tabbabarida AI-ga hawl gaar ah waxay sii ahaan doonaan kuwo diyaar u ah falanqayn ilaa moodeladooda aasaasiga ah laga dhigo kuwo duugoobay⁠(ku furmaa daaqad cusub). Jadwalka waqtiga oo dhammaystiran waxa uu ku qoran yahay halkan⁠(ku furmaa daaqad cusub).

Waxaa jira farsamooyin kala duwan⁠(ku furmaa daaqad cusub) oo ay horumariyeyaashu isticmaali karaan si ay u kordhiyaan waxqabadka moodelka, iyagoo isku dayaya inay yareeyaan dib u dhaca, hagaajiyaan saxnaanta, oo ayna yareeyaan kharashyada. Haddii ay tahay ballaarinta aqoonta moodelka iyadoo la adeegsanayo soo saaris lagu xoojiyay dib u soo helitaan (RAG), habeynta hab-dhaqanka moodelka iyadoo la adeegsanayo tabbabar dheeraad ah, ama dhisida moodel si gaar ah loo tababaray oo leh aqoon cusub oo ku saabsan qayb gaar ah, waxaan horumarinay xulashooyin kala duwan oo taageeraya hirgelinta AI ee macaamiisheena. Maanta, waxaan daahfuraynaa sifooyin cusub si horumariyeyaasha loo siiyo xakameyn dheeraad ah oo ku saabsan hawl ku tabbabarka AI-ga ee dheeraadka ah iyadoo la adeegsanayo API-ga, waxaanan soo bandhigaynaa siyaabo dheeraad ah oo loola shaqeeyo kooxdayada khubarada AI iyo cilmi-baarayaasha si loo dhiso moodello gaar ah.

We launched the self-serve fine-tuning API⁠(ku furmaa daaqad cusub) for GPT‑3.5 in August 2023. Since then, thousands of organizations have trained hundreds of thousands of models using our API. Fine-tuning can help models deeply understand content and augment a model’s existing knowledge and capabilities for a specific task. Our fine-tuning API also supports a larger volume of examples than can fit in a single prompt to achieve higher quality results while reducing cost and latency. Some of the common use cases of fine-tuning include training a model to generate better code in a particular programming language, to summarize text in a specific format, or to craft personalized content based on user behavior.

For example, Indeed⁠(ku furmaa daaqad cusub), a global job matching and hiring platform, wants to simplify the hiring process. As part of this, Indeed launched a feature that sends personalized recommendations to job seekers, highlighting relevant jobs based on their skills, experience, and preferences. They fine-tuned GPT‑3.5 Turbo to generate higher quality and more accurate explanations. As a result, Indeed was able to improve cost and latency by reducing the number of tokens in prompt by 80%. This let them scale from less than one million messages to job seekers per month to roughly 20 million.

Today, we’re introducing new features⁠(ku furmaa daaqad cusub) to give developers even more control over their fine-tuning jobs, including:

• Epoch-based Checkpoint Creation: Automatically produce one full fine-tuned model checkpoint during each training epoch, which reduces the need for subsequent retraining, especially in the cases of overfitting
• Comparative Playground: A new side-by-side Playground UI for comparing model quality and performance, allowing human evaluation of the outputs of multiple models or fine-tune snapshots against a single prompt
• Third-party Integration: Support for integrations with third-party platforms (starting with Weights and Biases⁠(ku furmaa daaqad cusub) this week) to let developers share detailed fine-tuning data to the rest of their stack
• Comprehensive Validation Metrics: The ability to compute metrics like loss and accuracy over the entire validation dataset instead of a sampled batch, providing better insight on model quality
• Hyperparameter Configuration: The ability to configure available hyperparameters from the Dashboard⁠(ku furmaa daaqad cusub) (rather than only through the API or SDK)
• Fine-Tuning Dashboard Improvements: Including the ability to configure hyperparameters, view more detailed training metrics, and rerun jobs from previous configurations

DevDay bishii Nofeembar ee la soo dhaafay, waxaan ku dhawaaqnay⁠ barnaamijka Custom Model ee loo naqshadeeyay in lagu tababaro laguna hagaajiyo noocyo meel gaar ah, iyadoo lala shaqeynayo koox u go'an oo cilmi-baarayaal OpenAI ah. Tan iyo markaas, waxaan la kulannay daraasiin macaamiil ah si aan u qiimeynno baahiyahooda nooc-gaarka ah, waxaana horumarinay barnaamijkeenna si aan si dheeraad ah ugu badinno waxqabadka.

Maanta, waxaan si rasmi ah ugu dhawaaqeynaa adeeggeenna assisted fine-tuning oo qayb ka ah barnaamijka Custom Model. Assisted fine-tuning waa dadaal iskaashi ah oo lala sameeyo kooxaheenna farsamo si looga faa'iideysto farsamooyin ka baxsan API-ga fine-tuning, sida hyperparameters dheeraad ah iyo habab kala duwan oo parameter efficient fine-tuning (PEFT) ah oo miisaan weyn leh. Waxay si gaar ah waxtar ugu leedahay ururrada u baahan taageero dejinta dhuumaha xogta tababarka ee wax-ku-oolka ah, nidaamyada qiimeynta, iyo xuduudaha iyo hababka loo gaar yeelay si loo kordhiyo waxqabadka nooc adeegsiga ama hawsha ay u baahan yihiin.

Tusaale ahaan, SK Telecom⁠(ku furmaa daaqad cusub), oo ah operator isgaarsiineed oo u adeega in ka badan 30 milyan oo macaamiil ah oo ku sugan Koonfur Kuuriya, waxay doonaysay inay habayso nooc si uu khabiir ugu noqdo aagga isgaarsiinta iyadoo diiradda hore la saarayo adeegga macaamiisha. Waxay la shaqeeyeen OpenAI si ay fine-tune ugu sameeyaan GPT‑4 si loo hagaajiyo waxqabadkiisa wada sheekaysiyada la xiriira isgaarsiinta ee afka Kuuriya. Muddo dhowr toddobaad ah, SKT iyo OpenAI waxay horseedeen horumar waxqabad oo la taaban karo oo ku saabsan hawlaha adeegga macaamiisha isgaarsiinta—koror 35% ah oo ku yimid tayada soo koobidda wada sheekaysiga, koror 33% ah oo ku yimid saxnaanta aqoonsiga ujeeddada, iyo koror dhibcaha qanacsanaanta ah oo ka kacay 3.6 ilaa 4.5 (5tii) marka la barbar dhigo nooca fine-tuned iyo GPT‑4.

In some cases, organizations need to train a purpose-built model from scratch that understands their business, industry, or domain. Fully custom-trained models imbue new knowledge from a specific domain by modifying key steps of the model training process using novel mid-training and post-training techniques. Organizations that see success with a fully custom-trained model often have large quantities of proprietary data—millions of examples or billions of tokens—that they want to use to teach the model new knowledge or complex, unique behaviors for highly specific use cases.

For example, Harvey⁠(ku furmaa daaqad cusub), an AI-native legal tool for attorneys, partnered with OpenAI to create a custom-trained large language model for case law⁠. While foundation models were strong at reasoning, they lacked the extensive knowledge of legal case history and other knowledge required for legal work. After testing out prompt engineering, RAG, and fine-tuning, Harvey worked with our team to add the depth of context needed to the model—the equivalent of 10 billion tokens worth of data. Our team modified every step of the model training process, from domain-specific mid-training to customizing post-training processes and incorporating expert attorney feedback. The resulting model achieved an 83% increase in factual responses and attorneys preferred the customized model’s outputs 97% of the time over GPT‑4.

Waxaan aaminsanahay in mustaqbalka, inta badan ururradu ay horumarin doonaan noocyo la habeeyey oo loogu shaqsiyeynayo warshaddooda, ganacsigooda, ama adeegsiga ay leeyihiin. Iyadoo ay jiraan farsamooyin kala duwan oo loo heli karo dhisidda nooc gaar ah, ururro cabbir kasta leh waxay horumarin karaan noocyo shakhsiyeeysan si ay uga helaan hirgelintooda AI saameyn macno leh oo gaar ah. Furuhu waa in si cad loo xaddido adeegsiga, loo naqshadeeyo loona hirgeliyo nidaamyada qiimeynta, loo doorto farsamooyinka saxda ah, lana diyaargarowdo in waqti ka dib lagu celceliyo si nooc u gaaro waxqabadka ugu fiican.

OpenAI, ururrada intooda badan waxay si degdeg ah u arki karaan natiijooyin la taaban karo iyagoo adeegsanaya API-ga fine-tuning ee is-adeegga ah. Urur kasta oo u baahan inuu si qoto dheer u fine-tune gareeyo noocyadiisa ama uu ku dhex geliyo aqoon cusub oo gaar u ah domain-ka nooc, barnaamijyadeenna Custom Model ayaa caawin kara.

Booqo dukumentiyada API-ga fine-tuning⁠(ku furmaa daaqad cusub) si aad u bilowdo fine-tuning noocyadeenna.