TokenMeter

LLM API pricing · live reference

What does your AI actually cost per month?

Stop guessing from pricing pages. Plug in your real workload and see every major model — Claude, GPT, Gemini, DeepSeek — ranked by what you'll actually pay.

Claude Opus 4.8$25.00/1M outClaude Sonnet 4.6$15.00/1M outClaude Haiku 4.5$5.00/1M outGPT-5.5$30.00/1M outGPT-5.4$15.00/1M outOpenAI o3$8.00/1M outGPT-4.1$8.00/1M outGPT-4.1 nano$0.400/1M outGemini 3.1 Pro$12.00/1M outGemini 2.5 Pro$10.00/1M outGemini 2.5 Flash$2.50/1M outDeepSeek V4 Flash$0.280/1M outDeepSeek V4 Pro$3.48/1M outGrok 4.3$2.50/1M outGrok 4.1 Fast$0.500/1M outClaude Opus 4.8$25.00/1M outClaude Sonnet 4.6$15.00/1M outClaude Haiku 4.5$5.00/1M outGPT-5.5$30.00/1M outGPT-5.4$15.00/1M outOpenAI o3$8.00/1M outGPT-4.1$8.00/1M outGPT-4.1 nano$0.400/1M outGemini 3.1 Pro$12.00/1M outGemini 2.5 Pro$10.00/1M outGemini 2.5 Flash$2.50/1M outDeepSeek V4 Flash$0.280/1M outDeepSeek V4 Pro$3.48/1M outGrok 4.3$2.50/1M outGrok 4.1 Fast$0.500/1M out

The cost calculator

Workload

per request
Cache hit rate60%

Share of input served from prompt cache. Only applies to models that offer cached pricing.

Cheapest for this workload

DeepSeek V4 Flash

$26.29

/ month

  1. 01
    DeepSeek V4 FlashDeepSeek
    $26.29
  2. 02
    GPT-4.1 nanoOpenAI
    $39.00
  3. 03
    Grok 4.1 FastxAI
    $62.25
  4. 04
    Gemini 2.5 FlashGoogle
    $161
  5. 05
    Claude Haiku 4.5Anthropic
    $345
  6. 06
    Grok 4.3xAI
    $356
  7. 07
    DeepSeek V4 ProDeepSeek
    $496
  8. 08
    Gemini 2.5 ProGoogle
    $648
  9. 09
    OpenAI o3OpenAI
    $780
  10. 10
    GPT-4.1OpenAI
    $780
  11. 11
    Gemini 3.1 ProGoogle
    $990
  12. 12
    Claude Sonnet 4.6Anthropic
    $1,036
  13. 13
    GPT-5.4OpenAI
    $1,238
  14. 14
    Claude Opus 4.8Anthropic
    $1,727
  15. 15
    GPT-5.5OpenAI
    $2,475

Monthly estimate = per-request cost × requests per day × 30. Cached input billed at each model's cache rate where offered. Figures are estimates — verify against official pricing before committing spend.

Head-to-head comparisons

The matchups people search for most.

Every price, one table

ModelProviderInput /1MOutput /1MCached /1MContext
Claude Opus 4.8Frontier
Anthropic$5.00$25.00$0.500200K
Claude Sonnet 4.6Balanced
Anthropic$3.00$15.00$0.300200K
Claude Haiku 4.5Fast
Anthropic$1.00$5.00$0.100200K
GPT-5.5Frontier
OpenAI$5.00$30.00400K
GPT-5.4Balanced
OpenAI$2.50$15.00400K
OpenAI o3Frontier
OpenAI$2.00$8.00200K
GPT-4.1Balanced
OpenAI$2.00$8.001M
GPT-4.1 nanoFast
OpenAI$0.100$0.4001M
Gemini 3.1 ProFrontier
Google$2.00$12.001M
Gemini 2.5 ProBalanced
Google$1.25$10.00$0.3101M
Gemini 2.5 FlashFast
Google$0.300$2.50$0.0751M
DeepSeek V4 FlashOpen
DeepSeek$0.140$0.280$0.014128K
DeepSeek V4 ProOpen
DeepSeek$1.74$3.48128K
Grok 4.3Balanced
xAI$1.25$2.50256K
Grok 4.1 FastFast
xAI$0.200$0.500256K

Verified 2026-06-21· Prices in USD per million tokens. Always confirm on the provider's page before committing spend.

Need savings this month?

Choose a $99 quick audit or a $299 agent cost leak review.

If your bill is already real, the calculator is only the first pass. Send rough usage numbers or a provider export and get a short written report: likely waste, cheaper-model swaps, cache opportunities, and the next experiment to run. If the cost comes from agent loops, RAG over-retrieval, model routing drift, retry storms, cache misses, or coding-agent tool calls, choose the $299 Cost Leak Review path.

24h

Reply target after written scope

No keys

Usage summaries are enough

$99/$299/$1k

Paid only after scope acceptance

Agent cost leak and emergency path

The $299 path is for one bounded agent cost leak. The $1,000 emergency sprint is for runaway LLM bills where agent loops, RAG fan-out, retry storms, cache misses, routing drift, or launch-pricing risk need a fast containment plan.

Best fit: teams spending at least a few hundred dollars per month on OpenAI, Anthropic, Gemini, DeepSeek, or router APIs. This is optimization guidance, not provider billing support.

No API keys or private prompts here. A usage export, rough numbers, or redacted trace summary is enough for the first pass.

TokenMeter Pro · founder waitlist

Lock the $19/mo founder rate before Pro ships.

Pro will turn the calculator into a live AI spend dashboard: connect read-only provider usage keys, see monthly spend across Claude and OpenAI first, get budget alerts, and find cheaper model swaps before the next bill lands.

01 Real spend by provider
02 Budget alerts
03 Cheaper-model swaps

Waitlist signups get the founder rate; no payment until the product is ready and you choose to subscribe.

No spam. One email when Pro is ready; no payment now.