Free tools Windows power users keep installed
One-click scans. No signup required.
To estimate Claude costs on Amazon Bedrock, multiply your expected input, output, prompt-cache write, and prompt-cache read tokens by the rates for your exact Claude model, AWS Region, service tier, and routing method. Then compare that forecast with AWS usage and billing data. There is no single price for “Claude on Bedrock”: rates and availability vary, and savings from caching, batch inference, or cross-Region routing depend on the model and workload.
What determines Claude’s cost on Bedrock?
Bedrock charges for model inference according to the model and the amount and type of token usage. Input tokens, output tokens, cache writes, and cache reads can have different rates. The applicable rate can also depend on the model version, AWS Region, service tier, and inference route. A workload that uses more than one model, tier, or route needs a separate calculation for each segment.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
MINISFORUM MS-02 Ultra Workstation Mini PC, Intel Core Ultra 9 285HX (24C/24T, up to 5.5GHz), PCIe... | $1,659.00 | Buy on Amazon |
| 2 |
|
GMKtec EVO-X2 AI Mini PC Ryzen Al Max+ 395 Superchip 128GB LPDDR5X 2TB SSD | $3,649.99 | Buy on Amazon |
As an Amazon Associate I earn from qualifying purchases.
Before estimating, identify the exact Claude model and version, the Region where the workload is billed, the endpoint or inference profile, and the service tier. Confirm that the model and features you intend to use are available for that combination. AWS pricing and model availability can change, so check the live Bedrock pricing and model documentation for your configuration rather than carrying over a rate from a different model or Region.
How to calculate an estimate
For each workload segment and billing period, use:
Estimated inference cost = (uncached input tokens × input rate) + (output tokens × output rate) + (cache-write tokens × cache-write rate) + (cache-read tokens × cache-read rate)
#1 Best Overall
- High-Performance AI Processor：The MS-02 Ultra features an Intel Core Ultra 9 285HX (24C/24T, up to 5.5 GHz, 13 TOPS NPU), delivering fast and efficient performance for AI inference, algorithm development, and media workloads. A PCIe x16 expansion slot supports desktop-class GPU upgrades for advanced model training and accelerated computing tasks. It's ideal for creators, engineers, and teams handling intensive parallel workloads.
- 4 × M.2 PCIe 4.0 + 4 × DDR5 SODIMM slots:Four DDR5 SODIMM slots support up to 256 GB of memory, while ECC helps maintain data integrity in mission-critical environments. Four PCIe 4.0 M.2 slots support up to 24 TB of storage, supporting RAID 0/1/5/10, combining high-speed performance with data protection. It allows for the creation of independent scratch disks, media libraries, and project drives, providing high-throughput for production workflows.
- PCIe & USB 4.0 v2: Up to three PCIe slots can be equipped, including a dual-slot x16 GPU. The main slot supports PCIe 5.0, meeting the needs of high-bandwidth creative and computing workloads. USB 4.0 v2 (80Gbps) supports high-bandwidth external storage and displays.
- Ultra-fast Networking: Wi-Fi 7 further enhances wireless performance with next-generation speeds and low-latency stability. Intelligent bandwidth switching optimizes throughput in different network environments, ensuring optimal performance for enterprise or local networks. Dual 25GbE ports (providing up to approximately 3.125 GB/s bandwidth, about 25 times faster than traditional 1GbE), enabling seamless large-scale file transfers and parallel computing. 10GbE and 2.5GbE ports, with support for Intel vPro technology, ensure enterprise-grade remote management and deployment flexibility.
- Server-grade thermal architecture: Utilizing a dedicated CPU/GPU airflow design, equipped with a 6-pipe dual-fan cooler, it maintains stable performance even under sustained loads, delivering up to 140W Turbo power while maintaining a 100W TDP, and operating with noise levels as low as 36 dB. An integrated 350W power supply ensures stable and reliable output for demanding computing tasks and fully loaded extended configurations.
If rates are quoted per million tokens, divide each token total by 1,000,000 before multiplying by its rate. Use the matching rate for the relevant model, Region, tier, and route. Sum the results for each segment. This estimates Claude model-token inference; an AWS bill may include other Bedrock features or AWS services beyond those token charges.
1. Measure representative input tokens
Use AWS Bedrock’s CountTokens API on representative prompts where the model and endpoint support it. AWS says CountTokens does not incur charges, and its result is model-specific. Count realistic requests, including the system prompt, conversation history, tool definitions, and retrieved documents that will actually be sent. If Bedrock Runtime CountTokens is not supported for a Claude model, AWS documents Anthropic’s count_tokens API on bedrock-mantle for those cases.
2. Estimate output tokens from real tasks
Output length depends on what users ask and how the model responds. Use observed response lengths from representative tasks, or forecast low, typical, and high scenarios if the workload is new. A configured maximum output-token limit is a ceiling, not a prediction that the model will generate that many tokens on each request.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
3. Apply current rates and expected request volume
Look up rates for the exact model, Region, tier, and route. Include cache token rates if caching is enabled and the batch rate if the model and workload are eligible for batch inference. Multiply per-request token estimates by the number of expected requests in the period, keeping separate totals for any traffic billed at different rates.
4. Reconcile the forecast with actual usage
After deployment, compare estimated token totals and cost with AWS Cost and Usage Reports (CUR) and model invocation logs. Account for input, output, cache-read, and cache-write usage, and match each category to the applicable tier and routing rate. CUR aggregates usage and cost rather than providing a line item for every request; use invocation logs to investigate individual prompts and reconcile at a compatible model and usage-type level.
Which token categories belong in the estimate?
| Usage | What to count | Why it matters |
|---|---|---|
| Uncached input | Prompt and context tokens that are not billed as cache reads or writes | Use the input rate for the selected model, Region, tier, and route. |
| Output | Tokens generated by Claude | Use the output rate; estimate from representative responses rather than the output limit. |
| Cache writes | Eligible input tokens written to a prompt cache | Writes may have a different rate from ordinary input, and can cost more on some models. |
| Cache reads | Eligible input tokens served from a prompt cache | Use the cache-read rate and count realized reads, not merely prompts that could have been cached. |
Do not count the same token volume in more than one category in the estimate. In billing, usage types distinguish token categories and can also identify model, tier, and routing; use those distinctions when matching actual usage to rates.
Ways to reduce Claude inference spend
Remove input tokens that do not help
Trim duplicated instructions, stale conversation history, irrelevant retrieved content, and oversized tool descriptions. After changing a prompt, recount a representative sample against the same model used in production and inspect output lengths as well. Shorter prompts are not automatically better if they remove context needed for a correct answer, so evaluate cost alongside task quality.
Test prompt caching for repeated prefixes
Caching can help when many requests reuse a long, stable prefix, such as a system prompt, tool definitions, or shared reference material. Keep reusable content unchanged and place request-specific content after it. Supported Claude models may offer explicit caching and implicit cache behavior; model-specific minimum prefix sizes and time-to-live rules apply.
Cache support does not guarantee a hit, and cache writes may cost more than standard input tokens. Measure cacheReadInputTokens and cacheWriteInputTokens in responses or the corresponding billing usage types, then compare the savings from reads with the cost of writes. Prompt caching is for supported on-demand models and is not supported by the batch inference API.
Consider batch inference for eligible asynchronous work
Batch inference may suit offline tasks such as independent classification or summarization when an immediate response is unnecessary. AWS pricing states that select foundation models are priced 50% below on-demand inference; that does not establish eligibility for every Claude model or workflow. Check the current model list and Region availability before applying a batch discount to a forecast.
Rank #2
- EVOLUTION RYZEN AI MAX+ 395 MINI PC - GMKtec EVO-X2 is the next evolution in AI mini PC Ryzen Strix Halo series. Thanks to AMD Simultaneous Multithreading (SMT) the core-count is effectively doubled, to 32 threads. Ryzen AI Max+ 395 has 64 MB of L3 cache and can boost up to 5.1 GHz, depending on the workload. The Ryzen AI Max+ 395 is currently rated as the "most powerful x86 APU" on the market for AI computing.
- AI NPU with XDNA 2 ARCHITECTURE - Powered by 16 “Zen 5” CPU cores, 50+ peak AI TOPS XDNA 2 NPU and a truly massive integrated GPU driven by 40 AMD RDNA 3.5 CUs, the Ryzen AI MAX+ 395 is a transformative upgrade and delivers a significant performance boost over the competition. The Ryzen AI Max+ 395 excels in consumer AI workloads like the llama.cpp-powered application: LM Studio. Shaping up to be the must-have app for client LLM workloads, LM Studio allows users to locally run the latest language model without any technical knowledge required and unleash their creativity and productivity.
- AMD RADEON 8090S iGPU GAMING PC - The AMD Radeon RX 8060S offers all 40 CUs with up to 2.9 GHz graphics clock and uses the new RDNA 3.5 architecture. The powerful iGPU is positioned between an RTX 4060 and 4070 laptop GPU and therefore enables gaming in FHD at maximum details in most demanding games. The 8060S can also utilize the full 128GB pool, which is perfect for running LLMs such as Deepseek 70B Q8, which runs comfortably on this machine.
- EIGHT CHANNEL LPDDR5X - LPDDR5X is a new ground breaking memory small form factor installed on-board. With blazing speeds up to to 8000MT/s, it runs 1.5x faster than the DDR5 SODIMMs; 90% better performance over DDR5 SODIMMs in video conferencing and photo editing; 30% better performance in productivity apps; 12% better performance in digital content workloads.
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-X2 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and dual USB 4 40Gbps Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
Batch jobs run asynchronously through S3 input and output. They do not support tool calling, structured output, or multi-turn client interactions. Those constraints can rule out batch for interactive or tool-dependent workloads even when it could reduce inference cost.
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 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteCompare service tiers and routing against workload needs
Bedrock offers Standard, Flex, Priority, and Reserved service tiers, but availability and prices depend on the model and endpoint. Flex is positioned for flexible, non-time-sensitive work; Priority carries a premium for faster responses; Reserved involves dedicated capacity and term conditions. Compare eligible choices using both cost and the workload’s latency, availability, capacity, and data-residency requirements.
For the documented Claude Sonnet 4.5 case, AWS describes global cross-Region inference as approximately 10% less expensive on input and output token prices than geographic cross-Region inference, using the source Region’s price. This is not a general discount for all Claude configurations. Cross-Region routing may also be unsuitable when processing must remain within a particular Region; verify current model support and governance requirements.
Evaluate Provisioned Throughput with measured utilization
Provisioned Throughput is an option when a workload needs predictable capacity, but it involves capacity and model-unit choices and may require a commitment. AWS directs customers to request pricing from their account team. Compare any quote with measured on-demand spend and expected utilization; provisioning is not automatically cheaper.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How to read billing data and attribute costs
AWS CUR 2.0 provides detailed Bedrock line items by token type and usage. Relevant categories include input, output, cache read, and cache write. Usage types also identify the model, service tier, and routing, which helps match usage to the correct rates. If using cost-allocation tags, activate them before expecting them in CUR or Cost Explorer; AWS notes that activated tags may take up to 24 hours to populate.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →CUR does not contain per-request line items. Use model invocation logs when you need to understand the prompts behind a cost change, then reconcile those logs with CUR at a compatible model and usage-type grain. Cost-allocation tags can help organize shared workloads, but they do not replace request-level investigation in invocation logs.
Illustrative AWS pricing examples are not universal rates
AWS’s cited public pricing table for Claude 3.5 Sonnet (Public Extended Access), effective December 1, 2025, lists $6.00 per million input tokens and $30.00 per million output tokens for the listed Regions; its listed batch rates are $3.00 and $15.00 per million tokens. For Claude 3.5 Sonnet v2, the cited table lists on-demand rates of $6.00 per million input and $30.00 per million output tokens, batch rates of $3.00 and $15.00, cache-write at $7.50, and cache-read at $0.60 per million tokens for its listed Regions.
These are named public-access examples, not current rates for every Claude model, Region, tier, or route. Confirm the live price for your exact configuration before using any rate in a forecast.
What you need for a personalized monthly estimate
A useful monthly forecast requires the exact Claude model and version, AWS Region, tier and inference profile, representative input and output token counts, expected request volume, and expected cache-read and cache-write behavior. If considering batch or Provisioned Throughput, include model eligibility and any account-specific quote or terms. When an estimate depends on assumptions, calculate low, typical, and high scenarios rather than presenting a single unsupported monthly figure.
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.
