Tools / LLM API Cost Calculator

LLM API Cost Calculator

Estimate your monthly LLM API bill across Anthropic, OpenAI, and Google models. Enter your monthly token volume; costs update as you type. Prices as of July 3, 2026.

Cached input share = the fraction of input tokens served from prompt cache (system prompts, few-shot examples, long documents reused across requests). Models without published cache pricing bill cached tokens at the full input rate here.

Model Input $/M Output $/M Monthly cost vs cheapest
Claude Haiku 4.5
Fastest Claude tier; 200K context
$1.00 $5.00
Claude Sonnet 5
Intro pricing $2/$10 through Aug 31, 2026; 1M context
$3.00 $15.00
Claude Opus 4.8
1M context, no long-context premium
$5.00 $25.00
Claude Fable 5
Frontier tier; 1M context
$10.00 $50.00
GPT-5.2 Chat
Chat-optimized variant
$0.88 $7.00
GPT-5.2
Cached input $0.175/M
$1.75 $14.00
GPT-5.2 Pro
Highest OpenAI tier
$10.50 $84.00
Gemini 3 Flash
Text input rate; audio input $1.00/M
$0.50 $3.00
Gemini 3.5 Flash
Launched May 19, 2026
$1.50 $9.00
Gemini 3.1 Pro
Prompts >200K tokens bill $4/$18
$2.00 $12.00

Assumptions and sources

  • Standard synchronous API rates. Batch APIs typically run 50% cheaper on all three providers.
  • Cache write premiums (e.g. Anthropic's 1.25x on cache creation) are not modeled; for high cache-hit workloads the read savings dominate.
  • Long-context surcharges (Gemini 3.1 Pro above 200K tokens) and intro discounts (Claude Sonnet 5 through Aug 2026) are noted per model but not applied.

Prices verified July 3, 2026 against: Anthropic pricing · OpenAI pricing · Gemini API pricing

Go deeper on LLM cost control

A calculator tells you the sticker price; the bigger wins come from architecture. Our guide to LLM FinOps strategies covers tiered model routing and semantic caching that cut bills 60-80%, multi-LLM routing shows a measured 44% saving with no quality drop, and LLM cost monitoring tools compares the platforms that attribute spend per user and model. For context-window spend specifically, see context window optimization.