Nomic Atlas
Nomic Atlas is an AI tool primarily focused on the structuring of unstructured datasets. It encompasses an array of media types, including ...
Last verified:
What is Nomic Atlas?
Nomic Atlas is the unified workspace for data analysts and data scientists to visualize, curate, and analyze complex multimodal datasets including PDFs, images, transcripts, and text. It transforms raw, messy unstructured data into trustworthy decisions by providing embedding-native data maps with clustering, filtering, and drill-down capabilities. The platform connects directly to raw data and embeddings, enabling visual exploration of millions of rows across different formats without requiring code-heavy workflows.
Key features include AI-powered exploration tools with multimodal vector search, cluster analysis and topic modeling that generates clustering overlays and topic summaries directly in embedding space, and Atlas Analyst—a native AI agent that provides a conversational interface for asking questions and analyzing complex datasets. The platform also offers data communication and reporting capabilities allowing teams to collaborate on analysis, share deep links to specific data views, and align data scientists, analysts, and business users around findings with a single visual source of truth.
Atlas is designed for data analysts, data scientists, AI development teams, and machine learning engineers who need to explore patterns across millions of multimodal data points, curate training data, analyze customer feedback, validate model performance, or perform data labeling and curation workflows. It supports organizations looking to operationalize large datasets and build AI-ready data layers for unstructured data, analytics, and AI workflows.
The platform enables users to search and filter multimodal data using natural language or embeddings, run vector search and semantic queries across embeddings, transcripts, PDFs, and more, generate clustering overlays and track how clusters evolve over time, identify latent themes and emerging trends without writing code, and automate data curation using the Atlas API with seamless integration into existing data stacks.
Nomic Atlas pricing
Pricing model: Freemium
Starter (Free): Free tier limited to 250k public data points. Plus: $10/month per seat, $0.05/k after 1M public data points, $1/10M embedding API tokens. Business: $125/seat/month, $0.1/k after 1M private data points, includes collaboration on maps, dataset & organization role-based access control. Enterprise: Unlimited data points, unlimited seats, and unlimited embedding API tokens with custom deployment options. All plans include access to the Developer API endpoints (Parse, Extract, Embed, Datasets) with usage tracked by text tokens, images, and dataset storage.
Nomic Atlas pros
- Visualizes millions of multimodal data points in interactive embedding-native maps
- Supports PDFs, images, transcripts, text, video, audio, and pure embeddings
- No-code topic modeling generates clustering overlays and topic summaries automatically
- Atlas Analyst AI agent provides conversational interface for dataset questions
- Multimodal vector search explores relationships across text, images, and documents
- Natural language search and filtering works without writing SQL or code
- Deep links to specific data views eliminate need for screenshots when sharing
- Automated ingestion of files, transcripts, PDFs reduces manual preprocessing
- Side-by-side comparisons of AI outputs and source data validate model performance
- Tracks cluster evolution over time or shifts relative to numeric metadata
- Identifies latent themes and emerging trends without any coding required
- Duplicate detection and outlier uncovering through adjustable similarity thresholds
- API automation enables seamless integration into existing data stacks
- Collaborative environment aligns data scientists, analysts, and business users
- Reduces manual effort to explore, clean, and manage unstructured data with AI
- Works with various embedding models both closed and open source
- Python API and LangChain classes available for programmatic access
- ISO 8601 date/timestamp format ensures reliable parsing for CSV uploads
Nomic Atlas cons
- Free tier limited to 250k public data points
- Adding members to organization only available for Business and Enterprise plans
- Dataset-level permissions independent from organization-wide settings only on Business & Enterprise
- Requires Python environment setup and API key for programmatic access
- Learning curve for new users unfamiliar with embedding visualization concepts
- efficiency varies significantly by data type and structure
- Requires extensive user exploration to understand all capabilities
- No predefined data models means users must define their own schemas
- Map building only occurs after calling create_index, not automatic
- Logging can be annoying and requires manual disable configuration
Frequently asked questions about Nomic Atlas
What is Nomic Atlas and what does it do?
Nomic Atlas is a platform for interacting with unstructured datasets of text, image, video, audio, and embeddings at scale. It is the unified workspace for data analysts and data scientists to visualize, curate, and analyze complex multimodal datasets. Atlas provides embedding-native data maps with clustering, filtering, and drill-down capabilities, enabling visual exploration of millions of rows across formats without code-heavy workflows.
How do I set up the Nomic package for Atlas?
Install the nomic library in your Python environment using pip install --upgrade nomic, then import it with import nomic. Get your API key at https://atlas.nomic.ai/cli-login after signing into Atlas in your browser. Log in to your Nomic account in Python with nomic.login(YOUR_API_TOKEN_HERE) or in the terminal with $ nomic login YOUR_API_TOKEN_HERE.
How do I create a new dataset in Atlas?
Call AtlasDataset with a dataset name using from nomic import AtlasDataset and dataset = AtlasDataset('my-new-dataset'). The dataset won't have data yet. To add data, use the add_data method with dataset.add_data(data) where data is a DataFrame or list of dicts. Your data map will only build when you call create_index on the dataset.
How do I add data to an existing dataset?
Call add_data on the AtlasDataset object. For example: from nomic import AtlasDataset, dataset = AtlasDataset('my-existing-dataset'), then dataset.add_data(data=[{'text': 'my third document'}, {'text': 'my fourth document'}]). After adding new data, call create_index on the dataset to update the map.
What format should dates and timestamps be in Atlas?
Atlas works best with dates and timestamps in ISO 8601 format: YYYY-MM-DD for dates and YYYY-MM-DDThh:mm:ssZ for timestamps. When using the Python SDK, pass Python date or datetime objects which Atlas automatically converts to Arrow-compatible timestamps. For CSV uploads, using ISO 8601 format ensures the most reliable parsing.
What file types does Atlas support?
Atlas supports multimodal data including PDFs, images, transcripts, text, video, audio, and pure embeddings. It can visualize and explore complex datasets across all these formats in one visual, collaborative environment. The platform automatically handles automated ingestion of files, transcripts, and PDFs.
How does topic modeling work in Atlas?
Atlas generates clustering overlays and topic summaries directly in your embedding space without writing any code. It identifies latent themes and emerging trends automatically. You can track how clusters evolve over time or shift relative to numeric metadata, providing insight into latent patterns in your unstructured data.
What is Atlas Analyst and how do I use it?
Atlas Analyst is a native AI agent that provides a conversational interface for asking questions and analyzing complex datasets. You can use it to run vector search and semantic queries across embeddings, transcripts, PDFs, and more. It brings any filtered view or sample into context for deep analysis and iteration without requiring code.
How do I share insights and data views with my team?
Collaborate on analysis of text, images, PDFs, and embeddings by sharing deep links to specific data views, filters, or annotations—no screenshots needed. Share links are generated by clicking Share in the top-right corner of the Atlas interface and copying your link. This aligns data scientists, analysts, and business users around findings with a single visual source of truth.
How do I disable logging in Nomic Atlas?
Nomic utilizes the loguru module for logging. To disable or change the logging level, include this snippet at the top of any script: from loguru import logger, import sys, logger.remove(0), logger.add(sys.stderr, level='ERROR', filter='nomic'). This removes the default logger and adds a new one with ERROR level for nomic.