TinyFish scores 71 on AI benchmark while waiving retrieval fees for enterprise workflows

TinyFish, a web infrastructure provider for AI agents, has removed retrieval charges for its search and fetch APIs within specific rate limits, according to MarkTechPost. This change affects how organizations calculate the total cost of AI agents that read and process live web content. The company argues that teams often underestimate how retrieved content drives model processing expenses.

The removal of retrieval fees applies to search requests capped at 30 per minute and 500 per hour, and fetch requests limited to 150 URLs per minute and 1,000 per day. While the initial data retrieval is free, the content passed into the AI model still incurs token-based inference costs. This structure shifts the financial burden from the API call to the model processing stage.

Cost structure and performance

On the Sep 28, 2026 Artificial Analysis snapshot, TinyFish Search (web) achieved a score of 71 at a cost of $9.30 per 1,000 benchmark tasks. This performance metric includes both model processing and search time, averaging 59.0 seconds per task. The service recorded the lowest total cost among the 12 configurations displayed on the leaderboard, despite having the second longest time per task.

The highest score on the same leaderboard was Perplexity Search (medium) at 80, while the lowest was TinyFish at 71. A no-search baseline scored 33. These figures highlight the trade-off between speed, accuracy, and cost when selecting infrastructure for AI agents that interact with live websites.

Supplier and operational considerations

Enterprises increasingly route tasks to the cheapest capable models, a trend highlighted by Stripe’s agreement to acquire OpenRouter, a gateway spanning over 400 models. Routing lowers the price per token but does not reduce the number of tokens a retrieved page contains. The size of each returned page directly impacts downstream inference costs, making page quality and size critical factors in budgeting.

TinyFish uses a web-native AI model designed for search, retrieval, and interaction with live websites. The model runs on TinyFish-managed infrastructure with a cloud browser fleet that renders pages and handles interactions. No local GPU is required, and access is available via REST, Python and TypeScript SDKs, CLI, or MCP.

Operational impact

Organizations should review how AI agents consume retrieved content, as free retrieval does not eliminate processing costs. Teams usually need to verify whether fetched pages contain excessive text that inflates token usage. Check whether current workflows filter content before sending it to the model. Internal stakeholders, including finance and engineering leads, need to be informed about the shift in cost structure. This ensures budgets account for downstream inference rather than just API calls.

Background: AI agents

An AI agent is a software system that perceives its environment, makes decisions, and takes actions to achieve specific goals without constant human input. Unlike passive tools, agents execute tasks, browse data, and interact with other systems autonomously.

Read the full guide: AI Agents Defined: Autonomy, Risk, and Real-World Use

What to do and how to stay safe: TinyFish

• Monitor the volume of text retrieved by AI agents to control downstream token costs.
• Review rate limits for search and fetch operations to ensure compliance with free tiers.
• Evaluate whether the current model configuration balances speed and cost for your tasks.
• Once the vendor provides an update, check for changes in page rendering or filtering capabilities.

Step-by-step guide: Implementing AI Usage Policies at Work: A Step-by-Step Framework

General security guidance from the InfoSec Ledger newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Does TinyFish charge for search and fetch requests?

No, TinyFish removes retrieval charges within specific rate limits, but downstream model processing costs still apply.

How did TinyFish perform on the Sep 28, 2026 benchmark?

TinyFish Search (web) scored 71 and cost $9.30 per 1,000 tasks, with 59.0 seconds per task.

What are the rate limits for free retrieval?

Search is limited to 30 requests per minute and 500 per hour; fetch is limited to 150 URLs per minute and 1,000 per day.