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

Nganyari tanggal 8 Mei 2026: OpenAI lagi mboko sithik mungkasi platform fine-tuning. Platform iki wis ora bisa diakses déning pangguna anyar, nanging pangguna platform fine-tuning sing wis ana bakal tetep bisa nggawe tugas pelatihan sajrone sawetara wulan mbésuk. Kabeh model sing wis di-fine-tune bakal tetep kasedhiya kanggo inferensi nganti model dhasaré ora didhukung maneh⁠(mbukak ing jendhela anyar). Linimasa lengkapé ana ing kéné⁠(mbukak ing jendhela anyar).

Ana macem-macem teknik⁠(mbukak ing jendhela anyar) sing bisa digunakake para pangembang kanggo ningkatake kinerja model kanggo nyuda latensi, ningkatake akurasi, lan nyuda biaya. Apa iku ngembangake kawruh model nganggo retrieval-augmented generation (RAG), ngatur prilaku model nganggo fine-tuning, utawa mbangun model sing dilatih khusus nganggo kawruh anyar sing spesifik kanggo domain, kita wis ngembangake macem-macem pilihan kanggo ndhukung implementasi AI para pelanggan kita. Dina iki, kita ngluncurake fitur anyar kanggo menehi pangembang kontrol luwih akeh marang fine-tuning nganggo API lan ngenalake luwih akeh cara kanggo makarya bareng karo tim ahli AI lan panaliti kita kanggo mbangun model khusus.

We launched the self-serve fine-tuning API⁠(mbukak ing jendhela anyar) 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(mbukak ing jendhela anyar), 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(mbukak ing jendhela anyar) 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(mbukak ing jendhela anyar) 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(mbukak ing jendhela anyar) (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

Ing DevDay November kepungkur, kami ngumumake⁠ program Model Kustom sing dirancang kanggo nglatih lan ngoptimalake model kanggo domain tartamtu, kanthi kemitraan karo klompok peneliti OpenAI sing khusus. Wiwit wektu kuwi, kami wis ketemu karo puluhan pelanggan kanggo ngevaluasi kabutuhan model kustomé lan ngembangake program kami supaya bisa luwih ngoptimalake performa.

Dina iki, kami resmi ngumumake penawaran fine-tuning kanthi bantuan minangka bagean saka program Model Kustom. Fine-tuning kanthi bantuan iku upaya kolaboratif karo tim teknis kami kanggo nggunakke teknik ngluwihi API fine-tuning, kayata hyperparameter tambahan lan macem-macem metode parameter efficient fine-tuning (PEFT) ing skala luwih gedhe. Iki migunani banget kanggo organisasi sing butuh dhukungan nyiyapake pipeline data pelatihan sing efisien, sistem evaluasi, lan parameter uga metode khusus kanggo ngoptimalake performa model kanggo kasus panggunaan utawa tugase.

Contone, SK Telecom⁠(mbukak ing jendhela anyar), operator telekomunikasi sing nglayani luwih saka 30 yuta pelanggan ing Korea Selatan, pengin ngustomisasi model supaya dadi ahli ing domain telekomunikasi kanthi fokus awal ing layanan pelanggan. Dheweke kerja bareng OpenAI kanggo nindakake fine-tuning GPT‑4 supaya ningkatake performane ing obrolan sing gegayutan karo telekomunikasi nganggo basa Korea. Sajrone pirang-pirang minggu, SKT lan OpenAI ngasilake peningkatan performa sing nyata ing tugas layanan pelanggan telekomunikasi—peningkatan 35% ing kualitas ringkesan obrolan, peningkatan 33% ing akurasi pangenalan intent, lan kenaikan skor kepuasan saka 3,6 dadi 4,5 (saka 5) nalika mbandhingake model sing wis di-fine-tune karo 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(mbukak ing jendhela anyar), 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.

Kami percaya yen ing mangsa ngarep, mayoritas gedhe organisasi bakal ngembangake model kustom sing dipersonalisasi kanggo industri, bisnis, utawa kasus panggunaané. Kanthi macem-macem teknik sing kasedhiya kanggo mbangun model kustom, organisasi saka kabeh ukuran bisa ngembangake model sing dipersonalisasi kanggo ngasilake dampak sing luwih migunani lan luwih spesifik saka implementasi AI. Kunciné yaiku nemtokake cakupan kasus panggunaan kanthi cetha, ngrancang lan ngetrapake sistem evaluasi, milih teknik sing pas, lan siyap kanggo terus iterasi saka wektu ke wektu supaya model bisa nggayuh performa optimal.

Karo OpenAI, umume organisasi bisa ndeleng asil sing migunani kanthi cepet liwat API fine-tuning swalayan. Kanggo organisasi apa wae sing butuh fine-tuning model luwih jero utawa nambah kawruh anyar sing spesifik domain menyang model, program Model Kustom kami bisa mbantu.

Kunjungi dokumen API fine-tuning⁠(mbukak ing jendhela anyar) kami kanggo miwiti fine-tuning model kami.