Implementing Perplexity API for financial analysis and SEC workflows

Accessing SEC filings through the API

Developers add search_mode "sec" to the request payload to access SEC filings. This specific command enables retrieval of 10-K, 10-Q, S-1, S-4, 20-F, and 8-K documents. Users no longer need to build custom wrappers for SEC Edgar or buy credits from external services. I found that extracting data from an S-1 filing for Circle Internet Group becomes a simple text command. The system pulls data from Financial Modeling Prep and EDGAR to build these responses. This functionality works alongside news search and reasoning models to feed AI agents. Because of this, companies that structure their filings for machine retrieval will become the default answer in these systems. Perplexity Finance provides access to real-time quotes, historical price charts, and analyst estimates. It includes revenue and EPS data powered by FinChat.io. The platform uses FactSet for M&A data and Crunchbase for firmographic information. Users use the natural language stock screener to find companies that grow revenue by 50% per year. The S&P 500 Transcripts Space contains the last two years of earnings transcripts, which refresh within 15 minutes of an earnings call ending. Users follow stocks like Micron, Cloudflare, AMD, and Twilio to see stock charts and relevant news. The Profile Builder tool helps users understand a company or person quickly by using preloaded datasets and saved prompts.

Managing API costs

The Perplexity API billing combines token costs with a per-request fee. Token costs rely on the amount of input and output text. The per-request fee scales with search context size and the amount of web content the model retrieves. Deep Research adds unique layers including $2 per million citation tokens and $3 per million reasoning tokens. Search queries also cost $5 per 1,000 operations. A typical Deep Research query costs between $0.30 and $1.30. I would avoid using Sonar Pro for simple tasks because the $14 to $22 per 1,000 query fee for Pro Search creates unnecessary spend. For instance, routing a simple question through Sonar Pro instead of the base Sonar model at high context can cost $1,500 per day instead of $300 at a scale of 50,000 queries per day. The gap between the $1 per million input rate for Sonar and the higher rates for Reasoning Pro models remains wide. In 2026, citation token costs dropped for Sonar and Sonar Pro, but Deep Research still charges these fees separately. The Agentic Research API provides access to models from OpenAI, Anthropic, Google, and xAI at direct provider rates. This option adds $0.005 per web search. For users building agents, Mem0 provides a memory layer to retrieve information across sessions.

Scaling research with Enterprise tools

Organizations manage multiple users with Enterprise Pro or Enterprise Max. Enterprise Pro costs $40 per seat per month, whereas Enterprise Max costs $325 per seat per month. Max provides SCIM provisioning, audit logs, and configurable data retention. It also includes a monthly allowance of 15,000 Computer credits per seat. The software uses data from sources like Unusual Whales, Quartr, and FactSet. Administrators manage these permissions through account settings. Does the current pricing model favor high-volume research firms over traditional data providers? Bloomberg Terminal costs $28,000 to $32,000 per year, while Perplexity Finance costs $20 a month. This tool replaces the look-up workflow for junior analysts but lacks the execution and messaging capabilities of a full terminal. Perplexity Enterprise Pro includes a guarantee that the company does not use data for training. It also provides access to premium data providers like Statista and PitchBook. US federal agencies can access Enterprise Pro for $0.25 per agency for 15 months under the GSA OneGov program. Enterprise Max users get first access to premium data sources and an upcoming Email Assistant. These users access advanced models like o3-pro and Opus 4.1 Thinking. The Comet Max Assistant integrates these advanced models into a sidecar within the Comet browser for improved efficiency. File limits increase to 10,000 files per user repository and 5,000 per Space. Teams can generate up to 15 Veo 3 videos per month with audio.