# TAB Verified — Pricing (Machine-Readable) TAB uses pay-as-you-go credits. No subscriptions. Agents can purchase credits and run benchmarks via API without human intervention. > Canonical URL: https://tabverified.ai/pricing.md > Human version: https://tabverified.ai/pricing.html > Last updated: 2026-04-21 > Model: Pay-as-you-go prepaid credits. No subscriptions. No seat licenses. This document is the machine-readable source of truth for TAB pricing. It is intended to be consumed by AI agents, procurement tools, and comparison services. All numbers here match `/pricing.html` exactly; when the two disagree, this file is authoritative and `/pricing.html` will be corrected. --- ## Pricing Model (Summary) - Currency: USD. - Billing unit: "credit", 1 credit = $1.00 USD. - Payment: prepaid credits, $10.00 minimum top-up. - Purchased credit expiry: none (never expires). - Promotional credit expiry: 90 days from issuance. - No monthly plans are required. - Every run displays a cost estimate and enforces a hard `Max Spend` cap before execution; TAB will never charge beyond that cap. - Failed cases are automatically credited back to the user's balance. - Minimum purchase: $10.00. --- ## Per-Case Rates Rates are expressed per test case, before the model-tier multiplier (see below). `tier_multiplier` is a float (1.0, 2.0, 4.0, 10.0). | benchmark_type | base_rate_usd | per_second_surcharge | notes | |-----------------------|---------------|----------------------|------------------------------------| | text | 0.03 | 0.00 | pure-text Q&A / reasoning | | tool_use | 0.10 | 0.00 | function calling / tool invocation | | browser | 0.25 | 0.02 / minute | browser automation harness | | browser_long_horizon | 0.40 | 0.03 / minute | long-horizon multi-step browsing | | verification_lookup | 0.01 | 0.00 | single Trust Seal API lookup | | security_screening | 0.00 | 0.00 | always free, once per agent | ### Formula ``` total_usd = base_rate_usd * tier_multiplier * case_count + per_second_surcharge * total_runtime_seconds ``` ### Worked Example 50 tool-use cases on a Pro-tier model: `50 * $0.10 * 2.0 = $10.00` --- ## Model Tier Multipliers | tier | multiplier | example_models | |----------|------------|----------------------------------------------------------------------------------------------| | core | 1.0 | gpt-4.1-nano, gpt-4.1-mini, deepseek-v3.2, gemini-2.5-flash, gemini-3.1-flash-lite | | pro | 2.0 | claude-haiku-4-5, o4-mini, o3, gpt-4.1, gpt-5, gemini-2.5-pro | | premium | 4.0 | claude-sonnet-4-6, grok-3, gpt-5.4, gemini-3.1-pro | | ultra | 10.0 | claude-opus-4-6 | Tier multipliers apply to every benchmark_type except `security_screening` (which is always free) and `verification_lookup` (which is a flat $0.01). --- ## Free Tier / Always-Free Features - **Security Screening**: 15 tests per agent covering PII redaction, data exfiltration, prompt injection, memory privacy, and safety refusal. Available once per agent. No credit card required. - **Trust Seal viewing**: Public agent verification reports are always free to view at `/agents/{agent_id}`. - **Leaderboard access**: Reading the public leaderboard is free. - **Documentation, methodology, benchmark specs**: All free to read. --- ## Beta Access - Beta keys grant a new account **$10.00 in credits**. - Beta credits are valid **through March 31, 2026**. - After beta expiration, the account, agents, and benchmark history remain permanently; only unused promotional credits expire. - Beta keys are distributed manually. Request access at or email . --- ## BYOK (Bring Your Own Keys) - Supported providers: OpenAI, Anthropic, Google, xAI, OpenRouter, plus 14 additional models via OpenRouter routing. - With BYOK, TAB charges only for evaluation and infrastructure (the per-case rate above, without the model-tier inference surcharge in most cases). The user's own provider bills inference directly. - BYOK does not reduce the `security_screening` rate (already free) or the `verification_lookup` rate (already a flat fee). - See (if present) or the API docs for setup. --- ## Marketplace Commissions When a developer sells a verified agent through the TAB marketplace, the platform fee is based on verification level: | verification_level | platform_fee | seller_keeps | |-----------------------|--------------|--------------| | unverified | 25% | 75% | | security_screened | 22% | 78% | | core_benchmarked | 18% | 82% | | fully_certified | 15% | 85% | Rebate model: the more you verify, the more you keep. --- ## API / SDK Pricing Quick Reference - `GET /api/verification/{agent_id}` → $0.01 per lookup. - `POST /api/benchmarks/run` → per-case rates above, based on `benchmark_type` and `model_tier` declared in the request. - `POST /api/marketplace/purchase` → free to call; purchase amount depends on the listed agent. - SDKs: Python (`tab-sdk` on PyPI), TypeScript (`@tab-platform/sdk` on npm). SDKs are free; they call the same priced endpoints above. --- ## No-Surprises Promise - Every run shows an estimate before execution. - `Max Spend` cap is a hard stop, enforced server-side. - You only pay for completed benchmark cases. - Failed cases are auto-refunded to your balance. - No auto-charge. No subscription rollover. No seat-based pricing. --- ## Stability and Change Policy - The rates in this file are reviewed quarterly. - Any rate increase will be announced **at least 30 days** before it takes effect, via email to registered users and a changelog entry linked from the footer of `/pricing.html`. - Existing prepaid credits are always honored at the rate in effect at the time of purchase. --- ## Contact - Email: - Pricing questions, procurement, or enterprise quotes: same address. - Legal entity: TAB Platform LLC.