OpenAI announced a partnership with Scale AI to provide enterprise support for fine-tuning models, according to a company blog post published 24 August 2023 [source]. The arrangement positions Scale as a preferred partner for organisations seeking assistance with custom model training on OpenAI's platform.

Under the partnership, Scale will offer consulting and implementation services to enterprises fine-tuning GPT-3.5 Turbo and other OpenAI models. Scale's role includes data preparation, model evaluation, and deployment support. OpenAI stated the collaboration aims to "help businesses unlock the full potential of fine-tuning" by combining Scale's data infrastructure with OpenAI's model capabilities.

The announcement follows OpenAI's broader push into enterprise fine-tuning services. Fine-tuning allows organisations to adapt foundation models to specific use cases by training on proprietary datasets. OpenAI made fine-tuning for GPT-3.5 Turbo generally available earlier in August 2023, lowering the barrier for custom model development.

Scale AI, founded in 2016, operates data labelling and machine learning infrastructure services. The company has previously worked with OpenAI on reinforcement learning from human feedback (RLHF) for ChatGPT and other models. The partnership formalises Scale's position in OpenAI's enterprise ecosystem, though OpenAI did not disclose financial terms or exclusivity arrangements.

The move reflects growing demand for customised AI deployments in regulated industries and specialised domains. Enterprises have sought fine-tuning capabilities to improve model accuracy on domain-specific tasks while maintaining control over training data. OpenAI's partnership model delegates implementation complexity to third parties while retaining control of the underlying model infrastructure.

No technical changes to OpenAI's fine-tuning API or model behaviour were announced as part of the partnership.