by AnthropicLLM APIs
Per-token access to Anthropic's Claude models through one API, paid from prepaid credits, with a HIPAA BAA you can sign in the console.
Claude models from Haiku 5.5 at $0.10 / $0.50 per 1M tokens up to Fable 5.1, with no training on API data and a BAA you can sign in the console.
The Anthropic API gives you Anthropic's current Claude models through one Messages API, each with a 1M-token context window and tool use, and official SDKs in seven languages. Every model's token price is on one public page, and batch and prompt caching discounts can be combined. The Commercial Terms rule out training on your API content, the DPA is part of the terms, and the privacy center lists SOC 2 Type II and ISO 27001 for the API. Teams that want Claude without running inference get a pay-as-you-go service, and the same models are also sold on Amazon Bedrock, Google Cloud and Microsoft Foundry.
Against 3 other products we reviewed in LLM APIs: what you pay for with each, the five access facts, the plan limits we recorded and each review's bottom line. The note under the table says how each one was checked.
| Metric | Anthropic API (Claude) | Mistral AI API | OpenAI API | Google Gemini API |
|---|---|---|---|---|
| What you're paying for | Worth the priceClaude models from Haiku 5.5 at $0.10 / $0.50 per 1M tokens up to Fable 5.1, with no training on API data and a BAA you can sign in the console. | Worth the priceLow per-token prices on small open-weight models, with EU or US processing you can choose yourself and spending caps per workspace. | Saves engineering timeOne API key for OpenAI's GPT-6 models and hosted tools, priced per token with published batch, flex and caching discounts. | Easy to adoptGoogle's Gemini models by the token in US dollars, with an ongoing free tier to start and batch, flex and caching discounts to scale. |
| Overall score | 7.2 | 7.4Best | 7.1 | 5.3 |
| Free option | Free trial | Free plan | –Not measured | Free plan |
| Card to start | –Not measured | No card needed | –Not measured | No card needed |
| Getting access | Self-serve | Self-serve | Self-serve | Self-serve |
| Starter priceCheapest paid plan | $0.10/1M input tokens | $0.10/1M input tokens | $0.10/1M input tokens | $0.25/1M input tokens |
| Pricing model | Usage-based | Usage-based | Usage-based | Usage-based |
| Pricing page | Public prices | Public prices | Public prices | Public prices |
| How you pay | Prepaid credits by card, expire after 1 year; invoicing through sales | Prepaid pay-as-you-go credits, optional auto recharge | Prepaid credits ($5 minimum), expire after 1 year | Prepaid credits ($5 minimum), expire after 12 months |
| Free option | One-time free credits for new users, amount not stated | Free mode, no card; allowance not stated publicly | No free allowance stated | Ongoing free tier on select models, no card; limits shown in AI Studio |
| Reference scenario (100M in + 20M out tokens/mo) | $20 on claude-haiku-5-5, computed | $12 on ministral-3b-2512, computed | $20 on gpt-6-luna, computed | $55 on gemini-3.1-flash-lite, computed |
| Batch, cache and other discounts | Batch 50% off; cache hits 90% off on Haiku 5.5 | Batch 50% off; cached input 90% off | Batch and Flex 50% off; cached input 90% off | Batch and Flex (preview) 50% off; cached input 90% off |
| Trained on API data | No (Commercial Terms) | Unless you opt out (Free mode and paid) | No, unless you opt in | Paid: no. Free tier: may be used to improve Google products |
| Default retention | Deleted within 30 days | 30 rolling days for abuse monitoring | Abuse logs up to 30 days | Abuse logs for a "limited period", days not stated |
| Zero data retention | By agreement, through sales | By approval, pay-as-you-go, stateless endpoints only | By approval, through sales | Not guaranteed: paid prompts are logged for abuse monitoring (docs point to Gemini Enterprise Agent Platform, formerly Vertex AI) |
| SOC 2 / ISO 27001 | Both, listed for the API | Both, stated in the help center | Both, covering the API | Not stated for this API |
| HIPAA BAA | Standard BAA signed in the Console, if eligible | Not stated | Standard BAA accepted in the API Platform, if eligible | Not stated for this API |
| Data regions | US-only inference at 1.1x; otherwise global | EU and US endpoints at +10% | 10 storage regions by approval; processing in US, EU, UAE | No region choice stated |
| Official SDKs | Python, TypeScript, C#, Go, Java, PHP, Ruby | Python, TypeScript | Python, TypeScript/JavaScript, Java (beta), Go (beta), C#, Ruby | Python, JavaScript/TypeScript, Go, Java, C# |
| Console SSO | Through sales | SAML on the Enterprise plan | Custom/Enterprise billing plan | Not stated (Google account sign-in) |
| Bottom line | From its public pages, the Anthropic API has a complete per-token price list, discounts that combine and strong data terms, with a self-serve HIPAA BAA. Region choice is US only, zero data retention and SSO go through sales, and a few pages disagree on billing details. | From its public pages, the Mistral AI API has the lowest reference price in this comparison, EU or US endpoints you can pick yourself and spending caps. Training use is an opt-out, rate limits sit behind sign-in, and ZDR and SSO need approval or the Enterprise plan. | From its public pages, the OpenAI API has a complete per-token price list, published batch, flex and caching discounts, and audit reports that cover the API. Stricter data controls, residency and SSO go through sales, and a few docs pages lag behind the price list. | From its public pages, the Gemini API pairs an ongoing free tier on current models with a detailed USD price list and published batch, flex and caching discounts. Certifications, zero data retention and data regions sit with Google Cloud's Agent Platform, and a few billing pages disagree. |
Evidence: Anthropic API (Claude): facts checked 2026-10-15 from its public sign-up and pricing pages, no completed hands-on run; Mistral AI API: facts checked 2026-10-15 from its public sign-up and pricing pages, no completed hands-on run; OpenAI API: facts checked 2026-10-15 from its public sign-up and pricing pages, no completed hands-on run; Google Gemini API: facts checked 2026-10-15 from its public sign-up and pricing pages, no completed hands-on run. Prices in USD as listed, excluding tax. "–" means not recorded.
Review from public pages
Written from Anthropic API (Claude)'s public pages (pricing, security or privacy, and docs) as read on Oct 15, 2026. We did not sign up, and no hands-on run is published yet.
We read Anthropic's API pricing page, the models overview and the Claude Haiku 5.5 page, the deprecations, rate limit, data residency, API and data retention, batch, prompt caching, structured outputs and SDK docs, the Commercial Terms and DPA, the privacy center articles on retention and certifications, the help center articles on billing, SSO and Covered Models, and the claude.com API and pricing pages on Oct 15, 2026; we did not sign up or call the API. The trust center (trust.anthropic.com) is a JavaScript-only page, so the certification list comes from the privacy center.
The price list is complete: every model has base input, cache write, cache hit and output prices per 1M tokens in USD, plus batch prices. Haiku 5.5 is the one model priced by prompt length, with higher rates above 100,000 tokens. The rate limit docs publish per-tier limits and monthly spend caps, and the data docs explain US-only inference, zero data retention and HIPAA readiness in detail.
A few pages need a question. The pricing page says billing follows monthly usage while the help center says usage is paid from prepaid credits; the docs say conversation content is not retained by default while the privacy center says API data is deleted within 30 days; the US-only 1.1x covers all token types in the docs but input and output on claude.com; and free code execution is 1,550 hours a month in the docs and 50 hours a day on claude.com. New users get free credits, but no live page states the amount.
Claude Haiku 5.5 costs $0.10 / $0.50 per 1M tokens for prompts up to 100,000 tokens, with a 1M-token context and tool use. Our scenario comes to $20 a month (computed), $10 with the Batch API.
vs Google Gemini API: Google Gemini API's scenario model, Gemini 3.1 Flash-Lite, comes to $55 for the same tokens; OpenAI API's GPT-6 Luna also comes to $20.
Batch is 50% off input and output, cache hits are 0.1x the input price on Haiku 5.5 (0.05x on Opus 5.5 and Sonnet 5.5, 0.025x on Fable 5.1), and the pricing page says the discounts can be combined.
Compared: DeepSeek API has no batch API; its discounts are cached input and off-peak rates.
The Commercial Terms say Anthropic may not train models on customer content, the DPA with standard contractual clauses is incorporated, and the privacy center says API inputs and outputs are deleted within 30 days.
vs Mistral AI API: Mistral AI API's paid use is covered by an opt-out training control, not excluded by default.
Eligible organizations review and sign the standard BAA in the Claude Console. The privacy center lists SOC 2 Type I and Type II, ISO 27001:2022 and ISO/IEC 42001:2023 for the API.
vs Google Gemini API: OpenAI API also has a self-serve standard BAA, but only after an established history of API usage; Google Gemini API states no BAA for its Developer API.
Official SDKs cover Python, TypeScript, C#, Go, Java, PHP and Ruby, plus a CLI. Rate limits are published per tier and model, structured outputs follow a JSON schema, and you can set your own spend limit.
vs Mistral AI API: Mistral AI API's official SDKs are Python and TypeScript only, and its rate-limit numbers are shown only after login.
The pricing page says new users get "a small amount of free credits" but gives no amount. After that, usage is paid from prepaid credits, which expire one year after purchase and are non-refundable.
Workaround: Buy small amounts with auto-reload rather than a large prepayment.
Requests can be pinned to US-only inference at 1.1x the price on Claude 4.6 and later models; the default is global routing. No EU or other region is stated for the first-party API.
Workaround: The pricing page notes regional endpoints on Amazon Bedrock and Google Cloud, at a 10% premium over their global endpoints.
Zero data retention is enabled per organization by the sales team and does not cover the Console or Managed Agents; Fable 5.1 requires 30-day retention. The Console SSO settings appear only after working with sales or completing a merge proposal.
Workaround: For health data, the self-serve HIPAA BAA applies; the docs say a HIPAA-enabled organization does not also need zero data retention.
Spend caps are $500 a month on Start, $1,000 on Build and $200,000 on Scale; at the cap, requests stop until the 1st of next month. New organizations may start in an Evaluation tier whose lower limits are not published.
Workaround: Request a tier increase in the Console before you reach the cap.
Anthropic bills per million tokens, with input and output priced separately, from prepaid credits bought by card; invoicing is arranged with sales. There is no plan fee. Our scenario on Claude Haiku 5.5 is 100 × $0.10 + 20 × $0.50 = $20 a month at standard rates; the Batch API halves it to $10, and US-only inference makes it $22 (computed). Costs climb with prompts over 100,000 tokens on Haiku 5.5 (5x, $0.50 / $2.50), with larger models (Sonnet 5.5 $2 / $10, Opus 5.5 $4 / $20, Fable 5.1 $10 / $50), fast mode on Opus 5.5 ($8 / $40), 1-hour cache writes at 2x input, and web search at $10 per 1,000 searches. Claude 4.7 and later use a tokenizer that the pricing page says produces about 30% more tokens for the same text, so compare per-token prices with care. Fees exclude taxes, and posted rate changes take effect 30 days after posting, or sooner once you receive notice.
Pay-as-you-go has no term: either party may end the Commercial Terms for convenience with notice (30 days from Anthropic). You own your outputs, and under the DPA Anthropic returns a copy of customer data on request and deletes it within 30 days of termination. Usage can be exported to CSV from the Console. Unused prepaid credits are non-refundable. The same models also run on Amazon Bedrock, Google Cloud and Microsoft Foundry.
From its public pages, the Anthropic API has a complete per-token price list, discounts that combine and strong data terms, with a self-serve HIPAA BAA. Region choice is US only, zero data retention and SSO go through sales, and a few pages disagree on billing details.
| Limit | As recorded |
|---|---|
| How you pay | Prepaid credits by card, expire after 1 year; invoicing through sales |
| Free option | One-time free credits for new users, amount not stated |
| Reference scenario (100M in + 20M out tokens/mo) | $20 on claude-haiku-5-5, computed |
| Batch, cache and other discounts | Batch 50% off; cache hits 90% off on Haiku 5.5 |
| Trained on API data | No (Commercial Terms) |
| Default retention | Deleted within 30 days |
| Zero data retention | By agreement, through sales |
| SOC 2 / ISO 27001 | Both, listed for the API |
| HIPAA BAA | Standard BAA signed in the Console, if eligible |
| Data regions | US-only inference at 1.1x; otherwise global |
| Official SDKs | Python, TypeScript, C#, Go, Java, PHP, Ruby |
| Console SSO | Through sales |
Read from the vendor's public pages on Oct 15, 2026. No hands-on run is published yet. How we check
We have not published a sign-up run for Anthropic API (Claude) yet. The access facts above were read from its sign-up and pricing pages on Oct 15, 2026.
We have not published API calls for Anthropic API (Claude) yet.
Anthropic API (Claude) lists 6 plans. The cheapest paid plan, Claude Haiku 5.5 (prompts up to 100,000 tokens), costs $0.10/1M input tokens.
Prices as listed, USD, excluding tax; observed 2026-10-15.
Source: platform.claude.com pricing page. Prices change; the observed date above is when our agent read them.
We read the pricing, model, data-usage, trust and docs pages of three LLM APIs on the same day and scored them on the same six indicators. What 100 million input and 20 million output tokens a month cost, how each one treats your data, and who should pick which.
