Where should we send your report?

We are committed to protecting your privacy. You can unsubscribe at any time. See our Privacy Policy. By clicking submit, you consent to allow Tech 42 to store and process the information above.

Estimates only. All pricing reflects AWS Bedrock on-demand rates for us-east-1 as of May 2026. AWS pricing changes without notice — actual costs will vary by region, pricing tier, usage patterns, and prompt caching. Always confirm current rates at aws.amazon.com/bedrock/pricing before making infrastructure or budget decisions.
Load an example:
Configure workloads & pick your models
Chat
Conversational AI, chatbots, virtual assistants
System prompt + history + user message. Typical: 500–4,000.
Model reply length. Typical: 200–800.
Building AI agents? Try the AgentCore Calculator →
Choose a model for chat workloads:
Enter your volumes above, then click a row to select your model
Selected
Scores only where officially published by the model provider. N/A = no published score. Sources: Anthropic, Amazon, Meta (May 2026). † Llama 4 Scout context window may vary by deployment. ‡ Claude Opus 4.7 uses a new tokenizer that may generate up to 35% more tokens for the same input — per-token rates are unchanged but actual cost per request may be higher. Verify pricing →
Add a custom or fine-tuned model
Model Name
Input $/1K tokens
Output $/1K tokens
Document Processing
Extraction, summarization, classification
A standard page is ~750–1,500 tokens.
Extraction: 500–1,500. Summarization: 200–500/page.
Full IDP pipeline? Try the Document Processing Calculator →
Choose a model for document processing:
Enter your volumes above, then click a row to select your model
Selected
Scores only where officially published by the model provider. N/A = no published score. Sources: Anthropic, Amazon, Meta (May 2026). † Llama 4 Scout context window may vary by deployment. ‡ Claude Opus 4.7 uses a new tokenizer that may generate up to 35% more tokens for the same input — per-token rates are unchanged but actual cost per request may be higher. Verify pricing →
Add a custom or fine-tuned model
Model Name
Input $/1K tokens
Output $/1K tokens
Image Analysis
Vision, OCR, chart extraction, visual inspection
×
Input tokens (image) = width × height ÷ 750. 1920×1080 ≈ 2,765 tokens; 1024×1024 ≈ 1,334 tokens.
Input tokens (text) = the instruction sent with each image. A simple prompt ("Extract all text") ≈ 50–100 tokens. A detailed system prompt with instructions ≈ 200–500 tokens. Both image and text tokens are billed at the input rate.
Output tokens = the model's response. Classification or yes/no: 20–100. Structured extraction: 200–500. Full OCR or description: 500–2,000+.
Choose a model for image analysis:
Enter your volumes above, then click a row to select your model
Selected
Scores only where officially published by the model provider. N/A = no published score. Vision column shows models with native image support. Sources: Anthropic, Amazon, Meta (May 2026). † Llama 4 Scout context window may vary by deployment. ‡ Claude Opus 4.7 uses a new tokenizer that may generate up to 35% more tokens for the same input — per-token rates are unchanged but actual cost per request may be higher. Verify pricing →
Add a custom or fine-tuned model
Model Name
Input $/1K tokens
Output $/1K tokens
Your cost estimate
Enable a workload above and select a model to see your cost estimate.

Costs estimated using us-east-1 on-demand pricing, May 2026. Actual costs vary by region, pricing tier, prompt caching, and usage patterns. Pricing data sourced from AWS Bedrock pricing page and official model provider documentation. Benchmark scores sourced from official model cards only — N/A means no published score, not a low score. TPM (Tokens per Minute) figures reflect default on-demand quotas and may differ in your account. Image token estimates use a 750 px/token approximation across models. Verify current pricing →

Cost Report Preview