Claude API pricing in 2026: the complete cost breakdown

The short version

  • Claude list prices per 1M tokens, July 2026: Fable 5 $10 / $50, Opus 4.8 $5 / $25, Sonnet 5 $2 / $10 (intro — rises to $3 / $15 on 1 September 2026), Haiku 4.5 $1 / $5.
  • The number that ruins spreadsheets: Anthropic's newer models — Opus 4.7 and later, Fable 5, Mythos 5 and Sonnet 5 — use a tokenizer that produces about 30% more tokens for the same text. An independent measurement found 1.20× on English prose and up to 1.47× on technical documentation.
  • Which makes Sonnet 5's discount partly an illusion. From 1 September it costs the same per token as Sonnet 4.6 — on roughly a third more tokens.
  • Prompt caching cuts input cost by 90% and pays for itself after a single read. It is the largest lever on this page, and it is one config flag.
  • Thinking tokens are billed as output even when you never see them. On Opus 4.8, thinking display defaults to omitted: you pay for reasoning that is never returned.

Claude's pricing page looks simple. Two numbers per model, input and output, dollars per million tokens. Then the first real invoice arrives, it does not match the arithmetic you did, and you lose an afternoon working out why.

This page explains the gap. Every figure below comes from Anthropic's own documentation, retrieved in July 2026, with the traps marked where they exist.

What Claude costs: every model, every column

Claude API pricing is per million tokens, billed separately for input and output. Cache reads cost a tenth of base input; cache writes cost more than base input. Batch halves everything. Here is the full current table.

Anthropic list prices, USD per 1M tokens — July 2026
ModelInputCache write (5 min)Cache write (1 hr)Cache readOutputBatch in / out
Claude Fable 5$10$12.50$20$1$50$5 / $25
Claude Mythos 5 (limited)$10$12.50$20$1$50$5 / $25
Claude Opus 4.8$5$6.25$10$0.50$25$2.50 / $12.50
Claude Opus 4.7 / 4.6 / 4.5$5$6.25$10$0.50$25$2.50 / $12.50
Claude Sonnet 5 (intro, to 31 Aug 2026)$2$2.50$4$0.20$10$1 / $5
Claude Sonnet 5 (from 1 Sep 2026)$3$3.75$6$0.30$15$1.50 / $7.50
Claude Sonnet 4.6 / 4.5$3$3.75$6$0.30$15$1.50 / $7.50
Claude Haiku 4.5$1$1.25$2$0.10$5$0.50 / $2.50
Claude Opus 4.1 (deprecated, retires 5 Aug 2026)$15$18.75$30$1.50$75$7.50 / $37.50

The ratios matter more than the prices

Opus 4.8 costs exactly Haiku 4.5 on both input and output. Fable 5 costs Opus 4.8 and 10× Haiku. Those multiples are the whole argument for routing: if a task can be done by Haiku, running it on Opus burns five dollars for every one it needed.

Claude output price per 1M tokens — list, July 2026 $50Fable 5 $25Opus 4.8 $15Sonnet 5from 1 Sep $10Sonnet 5intro $5Haiku 4.5 Opus is 5× Haiku. Fable 5 is 10×. The tier you pick is the biggest cost decision you make.
Claude output pricing. The spread inside Anthropic's own range is wider than the spread between vendors.

The 30% tokenizer tax nobody budgets for

Anthropic's newer models tokenize text more finely, so the same document costs more to send. Anthropic states it plainly: Opus 4.7 and later Opus models, Fable 5, Mythos 5 and Sonnet 5 use a newer tokenizer that "produces approximately 30% more tokens for the same text". Sonnet 4.6 and earlier use the old one.

Thirty percent is the headline. The distribution is what will actually hit you.

What an independent measurement found

A developer ran identical files through Anthropic's own count_tokens endpoint on Opus 4.6 and Opus 4.7. The ratio was not flat.

Token inflation on identical text, old tokenizer vs new (measured via Anthropic's count_tokens API)
Content typeOld tokenizerNew tokenizerRatio
Technical documentation (English)4787041.47×
A real CLAUDE.md file (5 KB)1,3992,0211.45×
Shell script1,0331,4361.39×
TypeScript source1,2081,6401.36×
Python source8641,1121.29×
English prose5086111.20×
Dense JSON13,93915,7061.13×
Numeric CSV5,0445,4141.07×
Japanese / Chinese prose856 / 779866 / 7891.01×
Same text, new tokenizer: how much more it costs you Anthropic says “approximately 30% more tokens”. Measured, technical English is nearly 50%. Technical docs (EN)1.47× CLAUDE.md file1.45× Shell script1.39× TypeScript1.36× Python1.29× English prose1.20× Dense JSON1.13× Numeric CSV1.07× Japanese / Chinese1.01× 1.00× (no change) 1.30× (Anthropic’s stated average)
The 30% figure is an average. Instruction files and technical documentation — the content in every agent’s stable prefix — inflate by 45–47%.

Read the top two rows again. Technical English — documentation, instruction files, system prompts — inflates by 45 to 47 percent, not thirty. That is precisely the content sitting in every agent's stable prefix. The 30% figure is an average across content types you probably do not send much of.

Why Sonnet 5's price cut is partly an illusion

Sonnet 5 launched at $2 / $10 against Sonnet 4.6's $3 / $15. That reads as a 33% cut. But Sonnet 5 runs the new tokenizer and Sonnet 4.6 does not, so the same text becomes roughly 30% more tokens. The intro pricing roughly cancels the tokenizer, leaving you close to flat.

Then on 1 September 2026 the intro pricing ends and Sonnet 5 moves to $3 / $15 — identical to Sonnet 4.6 per token, on 30% more tokens. On that date, for the same workload, Sonnet 5 becomes more expensive than the model it replaced. Put it in your calendar.

Prompt caching: 90% off input, and it pays for itself immediately

Prompt caching stores a stable prefix — system prompt, tool definitions, retrieved documents — so you are not charged full price to send it again. A cache read costs 0.1× base input, a 90% discount. Writing the cache costs 1.25× input on a five-minute TTL, or 2× on a one-hour TTL.

Anthropic's own break-even, verbatim: caching "pays off after just one cache read for the 5-minute duration (1.25x write), or after two cache reads for the 1-hour duration (2x write)". Reuse a prefix even twice and you are ahead.

Prompt cache economics on a stable prefix (5-minute TTL) Request 1 — write 1.25× base input price Request 2 — read 0.1× already break-even Requests 3, 4, 5… — read, read, read 90% saved on every one the 5-minute window refreshes free on each use Worked example — Opus 4.8, 40,000-token stable prefix Uncached: $0.200 per request  •  Cache read: $0.020  •  One-off write: $0.250 Ten uncached requests: $2.00. One write plus nine reads: $0.43. Identical output.
Caching is the single largest lever in Claude pricing. One config flag, 90% off input.

Caching also buys throughput

Cached reads do not count toward your input-tokens-per-minute limit on current models. Anthropic's own example: with a 2,000,000 ITPM limit and an 80% cache hit rate you can effectively process 10,000,000 input tokens per minute. Caching is a rate-limit lever as much as a cost lever.

Why your cache silently misses

Caching fails quietly. No error, no warning — just a bill that looks like caching was never switched on. These are the failure modes, straight from Anthropic's documentation.

The minimum cacheable prefix

A prefix shorter than the model's minimum is not cached, and no error is returned. The thresholds differ per model, which is the trap: 512 tokens for Fable 5 and Mythos 5; 1,024 for Opus 4.8, Sonnet 5, Sonnet 4.6 and Sonnet 4.5; 2,048 for Opus 4.7; and 4,096 for Opus 4.6, Opus 4.5 and Haiku 4.5. A 2,000-token system prompt caches on Opus 4.8 and does absolutely nothing on Haiku 4.5.

A breakpoint on a block that changes

Anthropic names this as a common mistake: put a cache breakpoint on a block containing a timestamp or the user's latest message and "you pay for a fresh cache write on every request and never get a read". You are then paying 1.25× input, forever, for nothing.

The 20-block lookback window

Cache reads only find entries that earlier requests actually wrote. In a growing conversation, once the breakpoint drifts more than twenty blocks past the last write, you get no hit and pay full input on the entire prefix.

What invalidates a cache

  • Changing tool definitions — invalidates everything downstream.
  • Toggling web search or citations — invalidates system prompt and messages.
  • Switching between fast mode and standard — invalidates system and message caches.
  • Changing tool_choice, adding or removing images, changing thinking settings — invalidates messages.
  • Non-deterministic JSON key ordering, common in Go and Swift — silently breaks the prefix match.
  • Switching model version — invalidates every cached prefix, and on a newer model the cold rewrite is 1.3–1.45× larger.

Workspace isolation, since February 2026

Since 5 February 2026, caches on the Claude API are isolated per workspace rather than per organisation. Teams running several workspaces silently lost hits they used to get. Bedrock and Google Cloud still keep org-level isolation.

Batch API: 50% off, and it stacks

The Batch API discounts input and output by 50%, and Anthropic confirms the discount stacks with prompt caching and data residency. If a workload tolerates delay — evaluations, backfills, nightly classification, document enrichment — running it synchronously doubles the bill for no benefit.

Two limits worth knowing: batch is unavailable with fast mode, and unavailable inside Managed Agents sessions.

You pay for thinking tokens you never see

Extended thinking is billed as output, at the full output rate. Anthropic is explicit: "You're charged for the full thinking tokens generated by the original request, not the summary tokens. The billed output token count will not match the count of tokens you see in the response."

It gets sharper. On Opus 4.8, thinking display defaults to omitted — the field comes back empty and you are billed for every token of reasoning that produced it. Anthropic's note on that setting: "You're still charged for the full thinking tokens. Omitting reduces latency, not cost."

The billed count is readable at usage.output_tokens_details.thinking_tokens. If you have never looked at that field, look today. It is frequently the largest line in an agent's bill.

Every surcharge and multiplier in one table

Claude pricing modifiers — July 2026
ModifierEffectApplies to
Cache write, 5-minute TTL1.25× inputAll active models
Cache write, 1-hour TTL2× inputAll active models
Cache read0.1× input (90% off)All active models
Batch API−50% input and outputAll models; not with fast mode or Managed Agents
Long context (1M window)No surcharge. A 900k-token request bills at the same per-token rate as a 9k oneFable 5, Mythos 5, Opus 4.8/4.7/4.6, Sonnet 5, Sonnet 4.6
Data residency (inference_geo: "us")1.1× on all token categoriesOpus 4.6, Sonnet 4.6 and later only
Regional endpoints on Bedrock / Vertex+10% over global endpointsSonnet 4.5, Haiku 4.5, Opus 4.5 and later
Fast mode (Opus 4.8)$10 / $50 — 2× standard, for up to 2.5× output speedResearch preview; not combinable with batch
Priority TierNo longer purchasable. Existing commitments run to contract end
Tokenizer~30% more tokens for the same textOpus 4.7+, Fable 5, Mythos 5, Sonnet 5

The long-context rule is genuinely generous

This is the one place Claude is unambiguously cheaper than its rivals. OpenAI doubles input price above 272k tokens. Google raises Gemini 3.1 Pro from $2 / $12 to $4 / $18 above a 200k prompt. Anthropic charges nothing extra for the full 1M window, and caching and batch discounts still apply across it. On long-context work, the "expensive" model is often the cheap one.

What the tools cost

Claude tool and feature pricing, billed on top of tokens
FeaturePriceWatch out for
Web search$10 per 1,000 searchesSearch results bill as input tokens in that turn and in every subsequent turn
Web fetchNo surchargeA 500 KB PDF is roughly 125,000 input tokens
Code execution1,550 free hours/month per org, then $0.05 per container-hourMinimum 5-minute billing. Attaching files bills container time even if the tool is never called
Managed AgentsTokens + $0.08 per session-hourBatch, fast mode and residency modifiers do not apply here
Computer useNo fee; adds ~466–499 tokens to the system promptScreenshots bill as image tokens
Any tool presentAdds 290–804 tokens of system prompt, depending on modelOpus 4.7 is heaviest at 675–804 tokens

Rate limits and spend caps

Anthropic replaced the old numbered Tier 1–4 system with Start, Build, Scale and Custom. Each carries a monthly spend cap that acts as a hard stop: hit it and API usage pauses until the next calendar month unless you request a raise.

Anthropic API tiers — spend caps and headline limits
TierMonthly spend capOpus 4.x RPM / ITPM / OTPMFable 5 RPM / ITPM / OTPM
Start$5001,000 / 2M / 400k1,000 / 500k / 100k
Build$1,0005,000 / 5M / 1M2,000 / 1.5M / 300k
Scale$200,00010,000 / 10M / 2M4,000 / 4M / 800k
CustomNo capNegotiated with the account team

Two details catch people out. The Opus limit is a combined bucket across Opus 4.8, 4.7, 4.6 and 4.5 — switching versions does not hand you a fresh allowance. And Sonnet 5 has its own bucket, separate from Sonnet 4.x, so migrating genuinely doubles your Sonnet-class headroom.

How do you move up a tier?

Anthropic no longer publishes a deposit ladder. The current documentation says only that organisations move up "over time as you use the API", and that you can request an increase from the Console limits page. Any article quoting a specific dollar threshold for Start → Build is repeating pre-June-2026 information.

Is there a free tier?

Not a real one. Anthropic's own wording: "New users receive a small amount of free credits to test the API." No dollar figure is published. The widely-repeated $5 figure appears only in third-party blogs, so treat it as unconfirmed.

Of the three major vendors only Google offers a genuine ongoing free API tier, and only on Flash-class models — Gemini 3.1 Pro is paid-only. If a free tier is what you need for prototyping, that is where it lives.

Buying Claude on AWS, Azure and Google Cloud

Claude is sold through three cloud marketplaces, and the billing mechanics differ from the direct API in ways worth understanding before procurement routes through them.

Claude Consumption Units on AWS and Microsoft Foundry

Claude Platform on AWS and Claude in Microsoft Foundry bill in Claude Consumption Units at $0.01 per CCU — 100 CCU equals $1.00. Anthropic rates your usage in dollars at standard rates, applies any negotiated discount, converts to CCUs and reports hourly. Your cloud bill shows a single CCU line item.

The consequences are real: billing is arrears-only, there are no prepaid credits, discounts show up as fewer CCUs metered rather than a cheaper CCU, and fast mode is unavailable. Organisations on Claude Platform on AWS are also pinned to the Start tier and do not auto-advance.

The 10% regional endpoint premium

On Amazon Bedrock and Google Vertex, regional and multi-region endpoints carry a 10% premium over global endpoints, for Sonnet 4.5, Haiku 4.5, Opus 4.5 and every model since. Bedrock also does not support automatic caching, and its minimum cacheable prefix for Fable 5 is 1,024 tokens rather than 512.

Buying Claude through an AI gateway

An AI gateway resells Claude through an OpenAI-compatible endpoint, so a single key and balance cover Claude, GPT and Gemini together. Prices vary more than people expect: in our own Fable 5 comparison the spread runs from 40% below Anthropic's list to 10% above it for the identical model.

That is the whole reason to check rather than assume. The side-by-side table sits on our Fable 5 API page and the full ranking on our comparison page; how gateways actually charge is covered in our guide to AI API gateways.

How Claude compares to GPT and Gemini

At list price Claude sits between the two. Opus 4.8 at $5 / $25 undercuts GPT-5.5 at $5 / $30 on output, and both are roughly double Gemini 3.1 Pro at $2 / $12.

Flagship list prices compared, USD per 1M tokens, July 2026
ModelInputCached inputOutputLong-context penalty
Claude Opus 4.8$5$0.50$25None — full 1M at standard rates
OpenAI GPT-5.5$5$0.50$30Input 2×, output 1.5× above 272k
Google Gemini 3.1 Pro$2$0.20$12$4 / $18 above a 200k prompt
Claude Haiku 4.5$1$0.10$5None
Google Gemini 3.1 Flash-Lite$0.25$0.025$1.50

But per-token parity is not per-task parity, and the tokenizer comes back here. Artificial Analysis measures the real cost of running the same evaluation suite on each model: Claude Opus 4.8 cost $4,011, GPT-5.5 cost $2,159 and Gemini 3.1 Pro cost $815. Opus and GPT-5.5 carry nearly the same sticker price and a 1.9× gap in what they actually cost to run. We pull that apart in our real-cost comparison.

How to actually cut a Claude bill

  1. Turn on prompt caching. 90% off input, break-even after one read. Nothing else here comes close.
  2. Move async work to Batch. 50% off, stacks with caching. Evaluations and backfills have no business running synchronously.
  3. Route by task, not by habit. Opus is 5× Haiku. Classification, extraction and routing do not need the flagship.
  4. Cap the thinking budget. The default is tens of thousands of output-billed tokens per request. Lower the effort level or set an explicit cap.
  5. Check your minimum cacheable prefix. On Haiku 4.5 it is 4,096 tokens — four times Opus. Short prefixes cache on one and not the other, silently.
  6. Read thinking_tokens. You are billed for reasoning you cannot see. Measure it before optimising anything else.
  7. Set a workspace spend cap. The tier cap is a hard stop, not a warning — and a runaway agent will find it. Our guide to what AI agents actually cost covers how that happens.

Common mistakes

  • Budgeting from list price without the tokenizer. On technical English the new tokenizer inflates by up to 47%, not 30%.
  • Assuming Sonnet 5 is cheaper than Sonnet 4.6. After 1 September 2026 it is the same price per token, on more tokens.
  • Putting a cache breakpoint on a block that changes. You pay a write every request and never get a read.
  • Forgetting that web-search results are re-billed. They land in input on every subsequent turn.
  • Leaving Opus as the default model. Anthropic names this as one of the top two causes of surprise bills, alongside long sessions that were never cleared.

FAQ

How much does the Claude API cost per 1M tokens?

As of July 2026: Claude Fable 5 is $10 input / $50 output, Opus 4.8 is $5 / $25, Sonnet 5 is $2 / $10 on intro pricing (rising to $3 / $15 on 1 September 2026), and Haiku 4.5 is $1 / $5. Cache reads cost a tenth of the input rate on every model.

Is Claude more expensive than GPT?

At list price Claude Opus 4.8 is slightly cheaper than GPT-5.5 on output ($25 vs $30) and identical on input. In practice Opus costs more to run because it emits more tokens: Artificial Analysis measured $4,011 to run its evaluation suite on Opus 4.8 against $2,159 on GPT-5.5.

Does Claude have a free API tier?

No permanent one. Anthropic says new users get "a small amount of free credits" but publishes no figure. Of the major vendors only Google offers an ongoing free API tier, and only on Flash-class models.

What is the 30% tokenizer change?

Anthropic's newer models — Opus 4.7 and later, Fable 5, Mythos 5 and Sonnet 5 — use a tokenizer producing about 30% more tokens for the same text. Independent measurement puts it at 1.20× on English prose and up to 1.47× on technical documentation. Sonnet 4.6 and earlier are unaffected.

How much does prompt caching save?

Cache reads cost 0.1× base input, a 90% discount. Writing the cache costs 1.25× on a five-minute TTL. Anthropic states caching pays off after a single read at that TTL, or two reads on the one-hour TTL.

Why is my cache not being hit?

The usual causes are a prefix shorter than the model's minimum (4,096 tokens on Haiku 4.5, 1,024 on Opus 4.8), a breakpoint on a block that changes every request, a changed tool definition, or the breakpoint drifting more than twenty blocks past the last write. None of these produce an error.

Are thinking tokens billed?

Yes, at the full output rate, including tokens you never receive. On Opus 4.8 the thinking display defaults to omitted — the field returns empty and you are still billed. Check usage.output_tokens_details.thinking_tokens.

Does Claude charge extra for long context?

No. Fable 5, Mythos 5, Opus 4.8/4.7/4.6, Sonnet 5 and Sonnet 4.6 include the full 1M-token window at standard pricing. A 900k-token request bills at the same per-token rate as a 9k one. OpenAI and Google both charge a premium above their thresholds.

What discount does the Batch API give?

Fifty percent off both input and output. It stacks with prompt caching and data residency, but is unavailable with fast mode or inside Managed Agents sessions.

How do I get higher rate limits?

Request an increase from the Console limits page, or contact sales. Anthropic no longer publishes spend thresholds for advancing between the Start, Build and Scale tiers.

What happens when I hit my spend cap?

API usage pauses until the next calendar month unless you request a raise. Caps are $500 on Start, $1,000 on Build and $200,000 on Scale. Custom has no cap.

Is Claude cheaper on AWS Bedrock?

Not inherently. Token rates track Anthropic's list price, and regional or multi-region endpoints carry a 10% premium over global ones. Bedrock also does not support automatic caching, and billing on Claude Platform on AWS is arrears-only in Claude Consumption Units at $0.01 each.

Should I buy Claude through a gateway instead?

It depends entirely on the gateway. In our own Fable 5 comparison, prices run from 40% below Anthropic's list to 10% above it for the identical model. A gateway is not automatically cheaper — check the landed price with top-up fees included.

Why did my Claude bill jump without my usage changing?

The usual suspects, in order: you switched to a model on the new tokenizer, your cache stopped being hit after a tool-definition or model change, extended thinking is consuming more output than you realise, or a long session was never cleared.

Is Priority Tier still available?

No. Anthropic states that Priority Tier capacity commitments are no longer available for purchase. Existing commitments run to the end of their contract.

The bottom line

Claude's list price is honest and easy to read. Its effective price is a function of three things the price table does not show: how your text tokenizes, whether your cache is actually being hit, and how many invisible thinking tokens the model burns before it answers.

Get those three right and Claude is competitive with anything on the market. On long-context work, where OpenAI doubles its input rate and Google raises its own, Claude's flat 1M-token pricing makes it the cheapest flagship available. Get them wrong and you will pay a premium for a model you could have run on Haiku.

If you do only one thing: turn on prompt caching, then go and read thinking_tokens across your last hundred requests. Almost everyone is surprised by what they find.

Sources

  • Anthropic — Claude Platform pricing documentation. Retrieved 14 July 2026. Primary source for the price table, multipliers, tool pricing, CCU mechanics, the tokenizer statement and the long-context rule.
  • Anthropic — Prompt caching documentation. Retrieved 14 July 2026. Cache multipliers, minimum prefixes, invalidation triggers, workspace isolation.
  • Anthropic — Extended thinking documentation. Retrieved 14 July 2026. Thinking-token billing and the omitted display.
  • Anthropic — Rate limits documentation. Retrieved 14 July 2026. Start/Build/Scale tiers, spend caps, cache-aware ITPM.
  • Anthropic — Model deprecations. Retrieved 14 July 2026.
  • Anthropic — Claude Fable 5 and Mythos 5 launch announcement, 9 June 2026.
  • Independent tokenizer measurement via Anthropic's count_tokens endpoint, April 2026.
  • Artificial Analysis — model pages for Claude Opus 4.8, GPT-5.5 and Gemini 3.1 Pro Preview, including cost to run the Intelligence Index. Retrieved 14 July 2026.
  • OpenAI — API pricing page. Retrieved 14 July 2026.
  • Google — Gemini API pricing page. Retrieved 14 July 2026.
  • Best AI Gateways — our own gateway price comparison for Claude Fable 5, July 2026.

About this article. Written by the Best AI Gateways research team. Every price and multiplier above was taken from Anthropic's own documentation on 14 July 2026. Where a figure is contested, unpublished, or comes from a third party, we say so rather than present it as fact.

Published 14 July 2026. Last updated 14 July 2026. Independent ranking. We may be rewarded for recommending the service we rate best and sending users to it — that reward pays for the research behind this comparison and never buys a ranking position, at no extra cost to you. Model names and trademarks belong to their respective owners. Pricing is set by each provider and can change — always verify before you build.

Best AI Gateways

Copyright © 2026 Best AI Gateways

Independent ranking. We may be rewarded for recommending the service we rate best and sending users to it — that reward pays for the research behind this comparison and never buys a ranking position, at no extra cost to you. Information is provided “as is” without warranty. See our Disclaimer & Terms.