OpenAI's First DevDay: GPT-4 Turbo Gets Cheaper and Longer—Another Variable for Translation Workflows

On November 6, 2023, OpenAI held its first developer conference (DevDay) in San Francisco, launching GPT-4 Turbo: a context window expanded to 128K tokens (roughly 300 pages), knowledge updated to April 2023, input token prices cut to one third of GPT-4's, and output prices halved. The event also introduced GPTs—user-built custom models—and the Assistants API.

For the translation industry, two parameters from this launch directly rewrite workflow economics. First, long context: a 128K window means hundred-page technical documents and entire product manual chapters can be processed in one pass, so terminology consistency and contextual coherence no longer depend on complex segmentation and stitching. Second, price: the sharp drop in input costs makes the hybrid flow of LLM pre-translation plus human review cost-effective across more content types, pushing the applicability boundary of machine translation post-editing (MTPE) further outward.

GPTs and the Assistants API opened another door: building custom translation assistants with client- and product-line-specific termbases and style guides went from concept to low-barrier engineering. Competition among language service providers began shifting from who uses AI to whose AI workflows embed deeper client knowledge.

Our judgment at the time was validated in subsequent practice: long-context models markedly improved first-pass quality on technical documentation and legal contracts, while customization greatly increased the reuse efficiency of client terminology assets. Every tool generation moves the optimal division of labor in human-machine collaboration—continuously tracking and re-engineering the process is a provider's basic discipline.