AI Pricing Changelog & Provider Audit Trail
Every rate change, prompt caching revision, and context window adjustment on TokenMath is recorded and logged here. Prices are actively verified against first-party provider documentation with per-model verification timestamps.
Audit History & Rate Modifications
Calibrated to active deepseek-v4-flash API specification with tiered peak/off-peak pricing ($0.44/$1.32 peak, $0.22/$0.66 off-peak).
Updated API ID to canonical gpt-6-astra and promoted lifecycle status to Generally Available (Flagship).
Updated canonical API model ID to mistral-small-2603 matching official March 2026 documentation.
Calibrated context window to native 256k (262,144 tokens) and prompt cache rate to $0.05/1M (90% discount).
Calibrated context window to native 256k (262,144 tokens) and prompt cache rate to $0.015/1M (90% discount).
Context caching rate calibrated to official promotional $0.075/1M tokens through Dec 31, 2026 (subsequently $0.15/1M from Jan 1, 2027) with active promotional cache storage fee of $0.50/1M tokens/hour through Dec 31, 2026 ($1.00/1M/hour standard thereafter).
Enterprise flagship calibration: $5.00 input / $25.00 output, 1.0M token context window.
Frontier tier price reduction: $2.00 input / $10.00 output.
Context window expanded to 1,000,000 tokens for autonomous agentic workflows.
Frontier reasoning calibration: $10.00 input / $50.00 output, cached input discount reduced to $1.00, context verified at 1.05M tokens.
Calibrated against current OpenAI pricing: $4.00 input / $20.00 output, $0.40 cached, 1.05M tokens context.
Balanced tier rate correction: $2.00 input / $12.00 output, $0.20 cached, 1.05M tokens context.
Lightweight tier calibration: 90% prompt caching discount ($0.02/M), 1.05M tokens context.
Legacy third-party aggregator feed drift (subsequently overridden by authoritative first-party provider calibration).
Legacy third-party aggregator feed drift (subsequently overridden by authoritative first-party provider calibration).
Legacy third-party aggregator feed drift (subsequently overridden by authoritative first-party provider calibration).
Legacy third-party aggregator feed drift (subsequently overridden by authoritative first-party provider calibration).
Legacy third-party aggregator feed drift (subsequently overridden by authoritative first-party provider calibration).
Legacy third-party aggregator feed drift (subsequently overridden by authoritative first-party provider calibration).
Legacy third-party aggregator feed drift (subsequently overridden by authoritative first-party provider calibration).
Legacy third-party aggregator feed drift (subsequently overridden by authoritative first-party provider calibration).
How TokenMath Audits Prices
Our ingestion daemon runs daily cron routines against provider documentation endpoints, OpenRouter models registry, and LiteLLM canonical feeds. Any candidate rate delta exceeding 0.01% triggers automated diff isolation and is cross-referenced with primary provider documentation before entering the canonical registry.