Provider pricing · Last verified 2026-08-16

Claude API Pricing: Opus 5, Sonnet 5 & Haiku 4.5 Rates

Short answer: Anthropic's Claude family costs $1.00–$5.00 per 1M input tokens and $5.00–$25.00 per 1M output tokens. Sonnet 5's introductory $2/$10 rate is now permanent — the planned September 2026 price increase was cancelled.

Model lineup

Claude pricing table

Standard first-party API text rates per 1,000,000 tokens, verified against Anthropic's official pricing documentation.

ModelInput / 1MOutput / 1MContext windowMax outputSource
Claude Opus 5
Flagship reasoning
$5.00$25.001,000,000128,000Anthropic
Claude Sonnet 5
Balanced mid-tier
$2.00$10.001,000,000128,000Anthropic
Claude Haiku 4.5Lowest
Budget / high-volume
$1.00$5.00200,00064,000Anthropic

Prices are informational estimates in USD per Anthropic-published token rates and may exclude caching, batch discounts, tools or taxes. Sonnet 5 rates re-verified 2026-08-16; Opus 5 and Haiku 4.5 rates last fully verified 2026-08-09. The Anthropic invoice is the source of truth.

Pricing mechanics

The Sonnet 5 price increase that never happened.

Anthropic launched Claude Sonnet 5 with introductory pricing of $2.00 / $10.00 per 1M input/output tokens and announced a move to $3.00 / $15.00 on September 1, 2026. In August 2026 that increase was cancelled: the introductory rate simply became the standard rate. If you budgeted for the higher number, your mid-tier Claude workloads are now 33% cheaper on input than planned.

One structural detail matters more than any rate change: Haiku 4.5 caps at a 200K-token context window, while Sonnet 5 and Opus 5 accept 1M. Long-document workloads cannot simply route to the cheapest Claude — check fit before you route.

Sonnet 5 standard rate$2.00 in / $10.00 out per 1M tokens — introductory pricing made permanent, Sep 1 increase cancelled.
Haiku 4.5 context limit200K tokens. Long transcripts and repos need Sonnet 5 or Opus 5 at 1M.
Check fit firstConfirm your document fits the window with the Context Window Checker.
Live calculator

Your workload on Claude, priced monthly

Pick a preset or enter per-request token volumes and daily requests. All three Claude models are priced side by side.

Presets:
Cheapest 30-day total
Most expensive 30-day total
ModelPer requestPer day30-day estimate
Standard rates: first-party Claude API pricing. Excluded: caching, batch, tools and taxes. Pricing data verified —
Choosing a tier

Which Claude model fits which workload?

Purely on arithmetic: at document-processing volumes (60K in / 1K out per request, 1K requests a day) Claude Haiku 4.5 runs about $1,950/month where Opus 5 costs $9,750 — a 5x spread for the same token volumes, with Sonnet 5 in between at $3,900. Haiku 4.5 handles classification, extraction and short-form tasks inside its 200K window; Sonnet 5 is the balanced default with the full 1M context; Opus 5 is reserved for the hardest reasoning where output quality justifies the premium.

"Best fit" here is a cost result, not a quality ranking — benchmark quality on your own prompts before routing production traffic. Compare Claude against the other providers on the LLM API Pricing Comparison page.

High-volume / simple tasksClaude Haiku 4.5 — $1.00/$5.00 per 1M, 200K context. Classification, extraction, short answers.
Mixed production workloadsClaude Sonnet 5 — $2.00/$10.00 per 1M, 1M context. The balanced default.
Hard reasoningClaude Opus 5 — $5.00/$25.00 per 1M. Route selectively; cap output with the API Cost Calculator.
Changelog

Recent Claude pricing changes

2026-08-16 — Sonnet 5 increase cancelledThe planned September 1, 2026 move from $2.00/$10.00 to $3.00/$15.00 per 1M tokens was cancelled; the introductory rate is now the standard rate. Source: Anthropic pricing docs.
Verified weeklyAll rates on this page are re-checked against Anthropic's official docs every week; the Last verified date at the top reflects the latest full check.
FAQ

Claude API pricing questions

How much does the Claude API cost?

Claude models cost $1.00–$5.00 per 1M input tokens and $5.00–$25.00 per 1M output tokens at standard rates: Claude Haiku 4.5 at $1.00/$5.00, Claude Sonnet 5 at $2.00/$10.00, and Claude Opus 5 at $5.00/$25.00.

What is the cheapest Claude API model?

Claude Haiku 4.5 is the cheapest Claude model at $1.00 per 1M input and $5.00 per 1M output tokens — 5x cheaper than the flagship Claude Opus 5 — though its context window is 200K tokens versus 1M on Sonnet 5 and Opus 5.

Did Claude Sonnet 5 pricing change in 2026?

Yes. Anthropic launched Sonnet 5 with introductory pricing of $2.00/$10.00 per 1M input/output tokens and had scheduled an increase to $3.00/$15.00 for September 1, 2026. That increase was cancelled in August 2026, and the $2.00/$10.00 rate is now the standard price.

Claude Opus 5 vs Sonnet 5: which should I use?

On cost alone, Sonnet 5 ($2.00/$10.00) is 2.5x cheaper than Opus 5 ($5.00/$25.00) on both input and output, with the same 1M-token context window. Opus 5 is the flagship for the hardest reasoning tasks; route traffic to it selectively and benchmark quality on your own prompts before committing.

How does Claude pricing compare to OpenAI and Gemini?

At flagship level, Claude Opus 5 ($5.00/$25.00) matches GPT-5.6 Sol on input and undercuts it on output ($25 vs $30). At mid tier, Sonnet 5 ($2.00/$10.00) beats GPT-5.6 Terra ($2.00/$12.00) on output. Budget models are cheaper elsewhere: GPT-5.6 Luna ($0.20/$1.20) and Gemini 3.5 Flash-Lite ($0.30/$2.50) both undercut Haiku 4.5 ($1.00/$5.00). See the LLM API Pricing Comparison page for the full table, or the OpenAI and Gemini deep dives.

Methodology

Where these numbers come from.

All rates are taken from Anthropic's official pricing documentation and re-verified weekly. The calculator multiplies your token volumes by published per-million-token rates; it does not estimate tokenization and does not apply caching or batch discounts.

Official source: Anthropic pricing. Always confirm the Anthropic invoice before making purchasing decisions.

Compare providersOpenAI vs Claude vs Gemini side by side: LLM API Pricing Comparison.
Head to headSame-workload comparisons: OpenAI vs Claude and Claude vs Gemini.
Measure your promptPaste real text into the AI Token Counter.
What a token isDefinition, conversion by content type, and why output costs 5–8x input: What Is an AI Token?
HTML token savingsCleaning scraped markup removes 47–68% of input tokens: HTML Token Savings.
1M tokens in words750,000 words, 1,500 pages — and why only 5 of 8 models accept it in one request: How Many Words Is 1M Tokens?
Trim the promptFind repeated instructions, markup and filler: Prompt Weight Analyzer
How we verifySources, weekly cadence and what our estimates exclude: Methodology · Pricing changelog.