Last updated: June 9, 2026
Important: read before relying on any figure from this site.
CostMyAI shows AI model pricing and analyzes AI usage to surface potential savings. All figures are derived from public provider pricing and public benchmark scores. They are provided for estimation and planning. AI provider pricing changes frequently, sometimes without public notice. Verify every figure directly with each provider's official pricing documentation before making any financial, technical, or business decision.
The prices shown across CostMyAI, including in the Analyzer, the calculators, and the model database, are derived from publicly available provider pricing pages and are synced approximately every 6 hours. A sync captures the price at a point in time. It is a planning reference, not a binding quote or an invoice.
Actual cost depends on factors a public per-token price does not capture: prompt caching, batch processing, volume tiers, reserved capacity, committed-use agreements, API version differences, regional pricing, and any contractual arrangement between an organization and a provider. CostMyAI does not see or account for these.
The Analyzer reads the usage export you provide and groups it into workloads. For each workload it compares your current model against alternatives using public benchmark scores for the relevant task type (for example AA Coding Index for coding, TAU2 for agentic, GPQA for reasoning, IFBench for extraction and classification, LCR for retrieval). Savings are computed purely from live public prices and benchmark math. CostMyAI applies no estimation, no shaping, and no discount assumptions.
A savings figure is only labeled "verified" when the alternative scores at or above your current model on the named benchmark for that task. Where the benchmark gap is within margin, the saving is shown but labeled as not quality-certified. Where no benchmark exists for the task, CostMyAI refuses to claim a saving rather than guess. CostMyAI never claims a saving it cannot support with a named public benchmark.
A "verified" label means the public benchmark supports the switch on quality. It is not a guarantee that the alternative will perform identically on your specific data, prompts, or constraints. You remain responsible for validating any switch before you make it.
Providers update pricing regularly. Prices have historically fallen as competition increases, but they can rise for specific tiers or change structure entirely. CostMyAI makes a best-effort attempt to keep data current and cannot guarantee any figure reflects the price you will be charged today or at deployment. Verify the current price on the provider's official page before any financial commitment.
This list is for convenience. URLs change. Always confirm you are reading the provider's current, official pricing.
The Analyzer works from usage metadata: spend amounts, token counts, model names, request volumes, and date ranges. It does not require, request, or process your prompts, your model completions, or any conversation content. Do not include prompt or response text in the file you upload. See the Privacy Policy for how uploaded files are handled and retained.
To the fullest extent permitted by law, CostMyAI and its operators disclaim all liability for any loss, cost, damage, or other harm arising from reliance on any pricing figure, cost estimate, savings figure, or model comparison shown on this site, including the Analyzer's verdicts and any unexpected provider charges that follow a decision informed by the Service. Use the site at your own risk and independently verify every figure before acting on it.