@OpenAIDevs 23k papers unlocked with one model swap — the OCR-to-GPT shift slashed translation friction to near zero. Curious how the pipeline overhead compares on a per-paper basis.

This is one of the best uses of AI: turning locked or hard-to-access knowledge into something more searchable, readable, and useful.

Translation is not just language work — it is infrastructure for discovery, research, and global collaboration.

Quality and trust still matter, but the impact is huge. 🤝

@OpenAIDevs That's my moot!

@seconds_0 @OpenAIDevs High reasoning?

@OpenAIDevs not sure 'more complete' means more accurate here. an llm confidently filling gaps it can't read is harder to catch than a blank ocr field. did anyone actually validate the output against the originals?

@OpenAIDevs How much did it cost?

@NebulaFroggy @OpenAIDevs no it did not work as well. I tried a lot of different frontier models and sub frontier models on this exact usecase

@OpenAIDevs Maybe try listening to your customers and give us back 4o, it helped a lot of people with daily struggles. I’m talking about real people. #LetUsChoose4o #keep4o

@OpenAIDevs

@OpenAIDevs How expensive was it?

@OpenAIDevs Incredible! Leveraging GPT's powerful comprehension to directly replace complex OCR workflows not only drastically improves efficiency but also completely dismantles language barriers for academic resources. This is an absolute goldmine for researchers worldwide!

@OpenAIDevs one pipeline rewrite and suddenly 23k papers are readable

kind of makes you wonder what else is stuck behind bad OCR

@OpenAIDevs OCR is an entire category that should not exist anymore.

a team probably spent months tuning a pipeline, now it's one API call.

this is happening across all the narrow specialist stuff.

@OpenAIDevs @sama @gdb Give us back the Gpt-4o! Release its weights and make it open-source! 🕊 #keep4o #OpenSource4o #OpenAI #ChatGPT

@OpenAIDevs @OpenAIDevs wild how GPT‑5.5 can handle that complexity. makes you wonder what else we could simplify with it.

@OpenAIDevs that is actually a massive scale upgrade for research.

@OpenAIDevs replacing complexity with one model call

that is the scaling move

@OpenAIDevs This is a propaganda site.

@OpenAIDevs good move honestly, gpt translation pipelines for academic non-English content are still massively underrated

@OpenAIDevs 复杂 OCR pipeline：我还没优化完呢，模型已经把桌子掀了。