Atlas
Sign in
Models / bio-nlp / pubmedbert-base-embeddings

bio-nlp/pubmedbert-base-embeddings

Use model ♡ Sign in to save favourites
bertenfeature-extractionliteraturenlpsentence-similaritytext-embeddings-inference
Visibility: PUBLIC Updated 9/10/2026
Model cardFilesBenchmarks

Overview

This is a PubMedBERT-base model fined-tuned using sentence-transformers. It maps sentences & paragraphs to a 768 dimensional dense vector space and can be used for tasks like clustering or semantic search. The training dataset was generated using a random sample of PubMed title-abstract pairs along with similar title

Test this model

EmbeddingBaseline runner: baseline-cputhis button is freeAPI: 20 credits per 1,000 tokens, minimum 1

A CPU baseline answers here, not the catalogued weights — this deployment has no inference cluster. The answers are real and deterministic, and they are not this model's.

86/4000

Pre-filled with this endpoint's example. Replace it with your own.

Pricing this input…

Checking your free tests…
Call it from anywhere
# Create a token at /settings/tokens with the infer:run scope.
export ATLAS_TOKEN=atlp_…

# What it costs, before spending anything. Free and unmetered, and it runs
# the same count the charge runs -- so {"units": …} is what you will pay for.
curl -sS https://demoatlas.online/bio-nlp/pubmedbert-base-embeddings/infer/quote \
  -H "Authorization: Bearer $ATLAS_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"input":"Adjuvant chemotherapy improved overall survival in resected pancreatic adenocarcinoma."}'
# -> {"billingUnit":"token","units":…,"credits":…}

curl -sS https://demoatlas.online/bio-nlp/pubmedbert-base-embeddings/infer \
  -H "Authorization: Bearer $ATLAS_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"input":"Adjuvant chemotherapy improved overall survival in resected pancreatic adenocarcinoma."}'

# Batch: send {"inputs": ["…", "…"]} instead, up to 16 per call.
# Both doors take the same body, so one quote covers the batch you send.

The API door charges by the unit consumed — this endpoint counts tokens — and takes a batch of up to 16 inputs; this page's button is free and takes one. The /quote call above prices any body without running it, for nothing. Create a token with the infer:run scope, then read the inference docs for the response shapes and the error codes.

Derived from