Claude API pricing is usage-based. You pay separately for input tokens and output tokens, with additional billing categories for prompt-cache writes, cache reads, long-context requests, tools and (when applicable) a cloud endpoint premium. There is no single monthly API price: your bill depends on the model, the amount of text processed, how much context is reused and whether requests run synchronously or through the Batch API.
The rates below are representative prices published by Anthropic and checked on October 19, 2026. Model prices, IDs, availability and discounts can change, so use the current Anthropic pricing and deprecation pages before committing to a production budget.
As an Amazon Associate I earn from qualifying purchases.
Claude API prices at a glance
Anthropic quotes API prices in MTok, meaning one million tokens. Input tokens are the text and other content sent to Claude; output tokens are the text Claude generates. They are priced separately, so a response-heavy workload costs more than a prompt-heavy workload on the same model.
| Model | Input price per MTok | Output price per MTok | Price status |
|---|---|---|---|
| Claude Opus 4.1 | $15 | $75 | Representative Anthropic rate checked October 19, 2026 |
| Claude Sonnet 4 | $3 | $15 | Representative Anthropic rate checked October 19, 2026 |
| Claude Haiku 3.5 | $0.80 | $4 | Representative Anthropic rate checked October 19, 2026 |
These are not subscription prices and do not mean that every Claude model has the same rate. Anthropic can introduce models, change prices or retire model IDs. Treat the table as a dated reference rather than a permanent price list.
#1 Best Overall
- [Ultra-Lightweight Ear Buds Designed for Small Ears] Each earbud weighs only 3.7g and features a compact, ergonomic in-ear design made especially for small ears. Secure, low-profile, and comfortable for workouts, all-day wear or overnight sleeping.The ultra-slim profile sits flush in your ear with no hard edges poking or rubbing, so you can even sleep on your side without discomfort.
- [Immersive Stereo Sound with TOZO OrigX Technology] TOZO OrigX tuning delivers clear vocals, balanced mids, and natural stereo sound for music, podcasts, and videos.
- [Long Battery Life for Daily Use] Get up to 7 hours of playtime on a single charge, with up to 32 hours total using the charging case—ideal for workdays, commuting, and extended listening sessions.
- [Bluetooth 5.3 & Stable Connection] Bluetooth 5.3 provides fast pairing, stable wireless performance, and reduced dropouts as you move around home or office.
- [Deep Bass with Clear Vocals] High-performance drivers produce punchy bass while keeping vocals clean and detailed for everyday listening.
What appears on a Claude API bill?
Base input and output tokens
Base input covers the tokens in your prompt, conversation history, attached text and other request content. Output covers the tokens Claude generates. A long system prompt sent on every request is therefore a recurring input cost, while a verbose answer increases output cost.
Prompt-cache writes and cache hits
Anthropic separates the cost of writing content to a prompt cache from the cost of reading cached content later. The documented multipliers are:
| Cache operation | Multiplier against base input rate | How to interpret it |
|---|---|---|
| Five-minute cache write | 1.25× | The initial write costs 125% of the model’s normal input rate. |
| One-hour cache write | 2× | The longer retention window costs twice the normal input rate for the write. |
| Cache hit or refresh | 0.1× | Reusing cached content costs one-tenth of the normal input rate. |
Cache pricing applies to the cached portion, not automatically to every token in a request. You still pay normal input and output rates for uncached content. Anthropic says prompt-caching discounts can be combined with Batch API discounts, subject to the applicable billing rules.
Recommended Free Tools
Long-context pricing
Some requests qualify for a higher long-context tier. For Sonnet 4’s documented 1-million-token context beta, when total input exceeds 200,000 tokens, the example rates are $6 per MTok of input and $22.50 per MTok of output. The entire request is billed at the higher tier once that threshold is crossed; it is not split into a cheap first 200,000 tokens and an expensive remainder.
Tools and server-side charges
Tool definitions, including schemas, and tool-result content count toward token usage. A server-side tool can also add its own fee. Anthropic’s pricing page gives web search as an example at $10 per 1,000 searches, in addition to the standard token charges for the request. If your application lets Claude call a tool repeatedly, count both the number of calls and the extra text exchanged with each call.
Rank #2
- 【IP68 Waterproof Multifunctinal Smart Watch】Torjalph fitness activity tracker smart watch is IP68 waterproof technology ,you can wear it when swimming,handwashing,it's rain-proof which means that you perform multiple exercise modes outside without worrying about sweat or rain corroding your fitness smart watch,and this watch equipped with a 230 mAh large capacity battery magnetic charging ,and this sports activity tracker watch can be used 5-7 days If fully charged.
- 【Larger 1.83'' Full Screen Watch & Multi-dial Selection】1.83 inch full HD color touch screen smart watch for men / women provides excellent 240*284 resolution image quality ,This Bluetooth smart watch offers a better touch experience and remarkable immersion and ,more than 200 watch faces for you to choose from App"Da Fit ", you can change the dial style according to your mood to wear keeping the world on your wrist
- 【Accurate Activity Tracker Smart Watch 】sports watch activity tracker uses sweat and dust resistant breathable silicone strap, our Fitness Tracker Smart Watch for Android iOS Phones supports 12 sports modes,walking,running,bicycle,swimming,badminton,basketball,football,rope skipping. This bluetooth calls smart watch for men women is lightweight and stylish, bringing more convenient and relaxing sports experience while also providing more protection.
- 【Your 24 Hour Personal Daily Health Monitor】Fitness watch for women men full touch screen IP67 waterproof bluetooth smartwatch for Android iOS Phones with heart rate blood pressure monitor multifunction Andriod smart watch with steps counter sleep monitor sports fitness watch, monitoring blood oxygen saturation allows you to understand your physical condition in real time to track fitness progress ,to know your health status in time.
- 【More Functions Bluetooth Call Smart Watch】This fitness activity smart watch full screen for iPhone and Android phones includes a bunch of new features including:The bluetooth smart watch can hang up the phone, receive and display WhatsAPP /Text /Facebook /Twitter /Messenger reminder.Support sedentary reminder, alarm clock, stopwatch, timer,weather information,music control, remote camera control compatible with Android 5.0 and above or iOS 9.0 and above smartphones.
How to calculate a Claude request
For a request without caching, long-context pricing, tool fees or an endpoint premium, use:
cost = (input_tokens / 1,000,000 × input_rate) + (output_tokens / 1,000,000 × output_rate)
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Use the rates for the selected model and keep input and output measurements separate. The result is an estimate; the Anthropic Console or the cloud provider handling your request is the authority for your invoice.
Example: a 100,000-input, 20,000-output request
On Sonnet 4, 100,000 input tokens cost $0.30 (0.1 × $3), and 20,000 output tokens cost $0.30 (0.02 × $15), for an estimated $0.60. The same token counts on Opus 4.1 cost $1.50 for input plus $1.50 for output, or $3.00. On Haiku 3.5 they cost $0.08 plus $0.08, or $0.16.
Example: adding a Batch discount
Anthropic’s Batch API provides asynchronous processing at a 50% discount on both input and output tokens. The $0.60 Sonnet example would therefore be about $0.30 when it qualifies for batch pricing and has no other charges. Batch is intended for work that does not need an immediate response; it is not a discount for ordinary synchronous calls.
Rank #3
- [Low Noise High Fidelity] New bee guitar cord with 22AWG oxygen-free-copper center conductors provides enhanced signal transfer, the PVC insulation material and the high-density stranded copper double shield provide noise rejection for quiet operation
- [Ideal for Choice] New bee 1/4 straight and right angle metal connectors, ensures a secure connection in any instrument, pedal, or amp jack. L shape great fro on stage guitar stands
- [Wide Compatible] Suitable for mixer, power amplifier, microphone, effector, decoder, equalizer, electoronic organ, electric guitar, electric bass and other electronic equipments. Widely used in the installation of smart home, karaoke speaker, professional stage audio, etc
- [Easy Disassemble] Making incidental repairs quick and easy! Very nice, flexible cable
- [Dont Hesite to Order] 24 hours customer services and professional technology team are standing by
Example: crossing Sonnet 4’s long-context threshold
Suppose a beta request contains 250,000 input tokens and generates 20,000 output tokens. Using the documented long-context example rates, the estimate is 0.25 × $6 ($1.50) plus 0.02 × $22.50 ($0.45), totaling $1.95. The higher rates apply to the whole request because input exceeds 200,000 tokens. At ordinary Sonnet 4 rates, the same token counts would have been $0.75 plus $0.30, or $1.05, so crossing the threshold materially changes the estimate.
When prompt caching saves money
Caching is most useful when a large, mostly unchanged prefix is sent repeatedly: a policy manual, codebase instructions, product catalog or long system prompt. The first write costs more than a normal input read, but later hits are charged at 0.1× the base input rate.
Illustrative Sonnet 4 cache calculation
For a 500,000-token prefix on Sonnet 4, a five-minute write is priced at 0.5 × ($3 × 1.25), or $1.875. A cache hit for that same 500,000-token prefix is 0.5 × ($3 × 0.1), or $0.15, before accounting for uncached prompt text and output. A one-hour write would be 0.5 × ($3 × 2), or $3.00.
Whether caching lowers your total cost depends on reuse. If the prefix is used only once, the write surcharge adds cost. If it is reused many times within the cache window, the low hit rate can outweigh that initial charge. Measure cache-hit ratios rather than assuming every large prompt benefits.
Batch API: lower cost for asynchronous jobs
The Batch API allows asynchronous processing of large volumes of requests with a 50% discount on both input and output tokens. It suits tasks such as overnight classification, document extraction, evaluation runs and bulk transformations where a response does not have to arrive immediately.
Rank #4
- Digital Family Calendar: BSIMB digital calendar wall touch screen puts an end to your chaotic life. Designed for busy families, the digital calendar keeps you organised and in step with your family. Assign different colours and tasks to each family member to make family scheduling easier and more intuitive.
- Easy Setup and Auto-Sync: Plug in, connect to WLAN and auto-syn your electric calendar in minutes. Digital wall calendar compatible with Google calendar, iCloud calendar, Outlook calendar, Cozi calendar and Yahoo calendar. Add events directly on the device or through the ecalendar app. Electronic calendar stays current whether you're at work or traveling—no manual updates needed.
- All-In-One Home Assistant: Smart calendar builds habits and encourage independence with personalised chores. Display daily meal plan to reduce the daily ‘What's for dinner?’ stress. To-do lists keep track of everything you want. Turning chores into valuable achievements with stars and rewards, making life easier and more enjoyable.
- Manage Anytime Anywhere: BSIMB electronic calendar for family stay connected with the powerful ecalendar app. Access and manage your digital wall calendar anytime, anywhere with free features like stars & rewards, photo screensaver, chore charts, meals plans, to-do lists etc. Electronic planner paid features like Magic Import tailor advanced tools for users with special needs.
- Stay in Touch with Loved Ones: Share photos or videos from your phone to electronic wall calendar with app, instantly update your life moment, and keep in touch with your family, especially those who are far away; you can also surprise your smart calendar digital wall by sending him a wish card; it's also great for keeping tech-illiterate elders in touch with the youngsters.
- Use normal synchronous calls for interactive chat, live agents and latency-sensitive actions.
- Use batches when you can queue requests, monitor completion and tolerate delayed results.
- Apply the discount to both token categories, then add any cache, tool or long-context charges that still apply.
- Anthropic states that batch and prompt-caching discounts can be combined, making repeated large prefixes a potential fit for queued workloads.
How model choice changes your bill
Price alone is not a sufficient selection rule. Compare the model’s capability requirements, expected latency, throughput, context needs and output length alongside its token rates.
| Decision factor | Question to answer | Cost impact |
|---|---|---|
| Input versus output mix | Does the application send huge prompts, generate long answers, or both? | Output rates can dominate even when prompts are small. |
| Quality requirement | Does the task need the strongest reasoning available, or is a lighter model adequate? | A lower-priced model can reduce both input and output spend if quality remains acceptable. |
| Repeated context | Is the same prefix sent many times? | Prompt caching can reduce the effective cost of repeated input. |
| Latency and throughput | Can requests wait in a queue? | Batch processing trades immediacy for a 50% token discount. |
| Context length | Will input exceed 200,000 tokens on the Sonnet 4 long-context beta? | The documented higher tier applies to the entire request after the threshold. |
| Lifecycle status | Is the model ID still supported? | A retirement can force migration work even if the replacement’s rate differs. |
For a reliable comparison, record input and output prices separately. A model with inexpensive input but expensive output may be a poor fit for applications that produce long reports.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Using Claude through Amazon Bedrock
Amazon Bedrock is an alternative route to managed Claude access. It can matter when your organization already standardizes on a cloud provider, needs particular regional routing or wants provider-specific governance. Feature support, model IDs, quotas and operational behavior can differ from direct Anthropic API access, so verify the exact model and features before migrating.
Anthropic documents that regional endpoints for Sonnet 4.5 and later carry a 10% premium over global endpoints. That premium is tied to the endpoint choice, not a universal surcharge on every Claude request. Cloud-provider billing, taxes and account terms may also affect the final amount, so compare the complete provider bill rather than copying a direct-API estimate.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Is there a Claude API free tier?
The cited Anthropic pricing material establishes token rates and discounts but does not establish a current free-tier allowance. Do not assume that a consumer Claude plan, a trial credit or a temporary promotion applies to API usage. Check the current account terms in the Anthropic Console for any credit, limit or eligibility rule attached to your account.
Model availability and retirement risk
Claude availability is not static. Anthropic’s deprecation page records retirements in 2026 for several Claude 3 and Claude 4 models and lists replacement models. A price article that omits the model ID and a “checked on” date can become misleading when an old ID stops accepting requests.
- Store the exact model ID in configuration rather than relying on an informal family name.
- Check the deprecation page before launching a long-lived integration.
- Budget a migration test: prompt behavior, output quality, latency, context limits and token rates can all change with a replacement.
- Keep a fallback model only when your application can tolerate its different capabilities and price.
A practical budgeting workflow
- Measure real traffic. Capture input and output token counts by endpoint, model and feature. Separate ordinary requests from tool calls and long-context requests.
- Choose a billing scenario. Mark which traffic is synchronous, which can be batched and which repeats a stable prefix suitable for caching.
- Apply the base formula. Multiply each token category by its model’s per-MTok rate and add the categories together.
- Add exceptions. Include cache-write and cache-hit rates, the Sonnet long-context tier where applicable, server-side tool fees and any Bedrock endpoint premium.
- Model peaks, not just averages. A small number of long outputs or threshold-crossing prompts can dominate a monthly average.
- Set controls. Enforce maximum output tokens, reject unexpectedly large prompts, cap tool calls and alert when spend or token counts exceed a baseline.
- Recheck before release. Confirm live prices, supported model IDs and endpoint rules on the relevant Anthropic or cloud-provider console.
Common cost mistakes and fixes
Estimating with character counts
Characters and words are not billing units. Use the token counts reported by the API or your provider’s usage records; tokenization varies with language, formatting and code.
Counting only the visible user message
System instructions, conversation history, tool schemas and tool results can all add input tokens. Include the full request payload when forecasting.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchAssuming a cache hit is free
Cache hits are discounted, not free. The uncached part of the request and the generated output still incur their normal charges.
Missing the long-context boundary
A prompt just over 200,000 input tokens in the documented Sonnet 4 beta can move the whole request to the higher tier. Track input size before submission and split or summarize content when the application permits.
Applying the Batch discount to interactive traffic
The 50% reduction is for asynchronous Batch API processing. It does not automatically reduce the price of ordinary real-time calls.
Using an obsolete model name
Retired models can cause failed requests or an unplanned migration. Pin supported IDs and monitor Anthropic’s deprecation notices.
Free tools Windows power users keep installed
One-click scans. No signup required.
Bottom line
At the representative rates checked October 19, 2026, Haiku 3.5 is the lowest-cost option in this comparison, Sonnet 4 is the middle ground, and Opus 4.1 costs substantially more—especially for generated output. Your effective price can fall through caching and Batch processing, or rise through long-context billing, tool charges and regional cloud endpoints. Build the estimate from measured input and output tokens, document every exception and recheck live pricing before deployment.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
