Is the Papers with Code API still working?

No. The paperswithcode.com/api/v1 endpoints stopped responding when Meta shut Papers with Code down in July 2025. The domain now redirects to Hugging Face. Any client still hitting the old API receives an HTTP 30x or DNS failure.

Where can I download the historical Papers with Code data?

The paperswithcode/paperswithcode-data GitHub repository preserves a snapshot of the public dataset (papers, methods, evaluations, datasets) as JSON dumps. It is no longer being updated, but is the canonical source for historical PWC data.

What is the closest live API replacement?

There is no drop-in replacement that mirrors the entire PWC schema. For paper metadata, Semantic Scholar provides a free API. For trending papers, Hugging Face Trending Papers RSS works. For benchmark leaderboards with verified scores, Codesota maintains live data — accessible directly via /benchmark/{dataset_id} pages and JSON-LD Dataset schema embedded on every benchmark page.

Codesota · Papers with Code · APIReference · April 2026

Reference

The Papers with Code API, and what replaced it.

paperswithcode.com/api/v1 stopped responding in July 2025 when Meta shut PWC down. There’s no drop-in successor. For each thing the old API gave you, here’s the closest live-and-maintained replacement.

§ 01 · API mapping

What the old endpoints gave you, and where to get it now.

Old endpoint	What it returned	Live replacement
/api/v1/papers/	Paper metadata (title, authors, arxiv_id, abstract, code links)	Semantic Scholar API ↗ · /papers
/api/v1/datasets/	Datasets with their metrics and modalities	/browse (per area / task / dataset)
/api/v1/evaluations/	Benchmark results (paper × dataset × metric × value)	/benchmark/<dataset_id> — every page emits Dataset JSON-LD
/api/v1/methods/	Models and their architectures	/models (paginated by area)
/trending	Trending papers feed	/papers-with-code/trending · Hugging Face Papers ↗
/api/v1/sota	SOTA per (task, dataset, metric)	/llm, /ocr, /code-generation — top score per benchmark, dated

§ 02 · Historical archive

The PWC dataset is preserved on GitHub.

paperswithcode/paperswithcode-data hosts the last public snapshot as JSON dumps — papers, methods, evaluations, datasets. It is no longer updated and the schema won’t change, but for backfilling old projects against the pre-July-2025 state of ML research it’s the canonical source.

If you need fresh, dated, reproduced scores for a specific benchmark — that is what Codesota maintains. Each /benchmark page includes a Dataset JSON-LD block you can scrape directly.

Three places to go from here.

Replacement

Papers with Code is dead — alternatives

What replaced PWC for each use case: LLMs, OCR, speech, vision, robotics.

Live feed