{"id":"papr-embed-code-1-4b","slug":"papr-embed-code-1-4b","name":"papr-embed-code-1-4b","orgId":"papr","family":"papr-embed-code-1-4b","releaseDate":null,"availability":"announced","openWeights":true,"license":null,"paramsTotal":4000000000,"paramsSource":"slug","contextWindow":null,"knowledgeCutoff":null,"modelCardUrl":"https://huggingface.co/papr-ai/papr-embed-code-1-4b","hfRepo":null,"modalities":["text"],"aliases":["papr-embed-code-1-4b","papr-ai/papr-embed-code-1-4b","papr embed code 1 4b","papr-embed-code-1 4b"],"aliasNote":null,"origin":"tracked","tracked":{"trackedSince":"2026-10-28","announced":"2026-10-28","blurb":"Papr's open text embedding model for code retrieval (natural language to code, code to code, SQL), made by further training Qwen3-Embedding-4B on code-retrieval data. Papr says it is flat to worse than its base on non-code retrieval. Weights on Hugging Face under Apache-2.0, and hosted on Papr's embeddings API.","api":"live","apiModelId":"papr-embed-code-1-4b","openWeights":"yes","params":{"text":"4B (fine-tune of Qwen3-Embedding-4B)","source":"Papr's Hugging Face model card"},"facts":[{"label":"Task","value":"Text embeddings for code retrieval; one dense vector per input, no schema input"},{"label":"Base model","value":"Qwen/Qwen3-Embedding-4B"},{"label":"Embedding size","value":"2,560 dimensions; last-token pooling, L2-normalized; up to 8,192 tokens"},{"label":"Weights","value":"Hugging Face papr-ai/papr-embed-code-1-4b, fp32, Apache-2.0 (repo created Oct 19, card added Oct 21)"},{"label":"API","value":"POST https://memory.papr.ai/v1/embeddings with model papr-embed-code-1-4b"},{"label":"Languages","value":"English and code only"},{"label":"Training data note","value":"Trained on the train splits of 8 of the 10 CoIR tasks and SciFact, so CoIR and SciFact are mostly in-domain, per Papr"}],"claims":["Lab claim: CoIR mean nDCG@10 of 80.13 over 10 tasks (mostly in-domain, per Papr), run with stock mteb 2.10 in fp32.","Lab claim: on held-out retrieval tasks it trails Qwen3-Embedding-4B on ArguAna, SciDocs, FiQA and TREC-COVID (87.04 vs 93.00) and gains on NFCorpus and CodeEditSearch."],"sources":[{"kind":"repo","title":"papr-ai/papr-embed-code-1-4b on Hugging Face","url":"https://huggingface.co/papr-ai/papr-embed-code-1-4b"},{"kind":"repo","title":"Papr-ai/papr-embed-evals on GitHub","url":"https://github.com/Papr-ai/papr-embed-evals"},{"kind":"press","title":"Shawkat Kabbara on X: papr-embed release article","url":"https://x.com/shawkatkabbara/status/2115476518088716604","date":"2026-10-28"}],"firstBenchmark":null},"organization":{"id":"papr","name":"Papr","slug":"papr","country":null,"url":"https://papr.ai","color":"#11110F","mark":"P"},"price":null,"answer":{"word":"Pending benchmarks.","scope":"Announced Oct 28, 2026; no independent results yet.","lead":"As of Nov 3, 2026, papr-embed-code-1-4b is pending benchmarks on isitsota: announced Oct 28, 2026, with no independent results on the boards we track yet.","confidence":"low"},"status":"PENDING_BENCHMARKS","verdicts":[]}