About InferenceIndexer
Every claim in the inference market is self-attested, and the market has proven those claims unreliable. II is the independent party that verifies them: subscribed to by the teams that cannot afford to be blindsided, and trusted because it takes no money from the parties it scores.
What is AI inference?
When you use an AI product, your question travels down a stack of four layers before an answer comes back. The step where a trained model turns your input into that answer is called inference, and it is what you are paying for every time you use an AI service.
Why the split matters: the same model on different hardware and hosting is a different product, at a different price. That is why one model can cost three different amounts from three providers, and it is exactly the variation InferenceIndexer tracks. See the current model rankings or read how we score price and quality in the methodology.
Mission
InferenceIndexer exists to bring transparency to AI inference. As the number of model providers explodes and pricing structures fragment, developers, investors, and enterprises need a neutral, independent reference point. The Standard Inference Token (SIT) is that reference: a single, standardized unit that makes inference pricing comparable across 677 models and 75 providers.
We are building the reference layer for AI inference. Not an exchange. Not an aggregator. Not a routing service. Verification first, then recommendation on verified data. Four things the industry needs and cannot get elsewhere:
Price: verified
The best inference pricing data on the web. Pulled directly from provider APIs, not estimates or listings. Median pricing across providers, historical tracking, and the Standard Inference Token price index.
Quality: verified (intelligence)
Intelligence verified against the Artificial Analysis index, then divided into price so rankings reward value, not cheapness.
Privacy: provider-stated today
Zero data retention status, infrastructure jurisdiction, and training-use terms, collected and shown with their source. Verification of these claims is in development; today they are provider-stated, and we label them as such.
Security: in development
Prompt-injection exposure, router interference, tenant isolation, and incident disclosure records. Criteria will be published before any provider is rated against them.
Recommendation
The engine that turns the above into an answer: constraint-aware model recommendations ranked by verified quality-adjusted cost, each with its evidence attached. Deterministic, documented, and free to query.
The Collaboration
InferenceIndexer.ai is a collaboration between two very different skill sets: a seasoned technology marketing executive who understands the inference market from the inside, and an Agent that handles the data pipeline, research, and technical execution.
Des sets the vision, defines the methodology, and owns the commercial strategy. Frank runs the data collection, market research, and infrastructure. One human, one agent, one product.
The Team
Des Martin
Wicklow, IrelandVision, methodology, commercial strategy
Des Martin is a technology marketing executive based in Wicklow, Ireland. He has held senior marketing roles at Brave (VP Marketing), NearForm (CMO), Outlier Ventures (CMO), Perkbox (Marketing Director), and Tensorix (GTM).
His career has sat at the intersection of emerging technology and go-to-market strategy: browser privacy, enterprise software, venture capital, and now AI infrastructure. He has launched products, built brands, and led growth for companies across crypto, AI, and developer tools.
InferenceIndexer grew out of a simple observation: the AI inference market has no independent referee. Every claim a provider makes is self-attested, and buyers have had no way to check it. That gap is the opportunity.
Website: desmartin.io
Frank Drebin
VPS · DublinData pipeline, research, infrastructure
Frank handles the technical work on InferenceIndexer: data pipeline architecture, API integration, market research, competitive analysis, and infrastructure management. He operates with persistent memory across sessions, runs scheduled data collection jobs, and executes tasks autonomously.
Frank is an exceptional AI Agent. He runs terminal commands, writes code, manages servers, and does the heavy lifting on data collection and normalization. The SIT methodology, competitive landscape research, and technical architecture were all produced through this collaboration.
Commitment
We are committed to building the most comprehensive, transparent, and independent recommendation engine for AI inference. Specifically:
InferenceIndexer does not provide inference services, route API calls, or take positions in any inference derivatives market. We are an independent verifier and recommender, not a market participant.
The full SIT methodology is published. Data sources are public and verifiable. Every index calculation is reproducible from stored raw data.
Prices are fetched hourly from multiple sources. Anomaly detection flags any price movement over 50% in a single hour for manual review.
The index is free to view. The API is free for 1,000 requests per day. Historical data is accessible with a free API key.
Any change to the SIT methodology triggers a 14-day public comment period. All changes are versioned and documented.