Anthropic API
Token-metered API access to Claude Opus, Sonnet and Haiku
Overview
The Anthropic API is direct, usage-billed access to the Claude model family, with prompt caching, batch discounts, native tool use and a 1M-token context option on the top-tier models.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 11 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- Lowest latency in the lineup
- Prompt caching + batch discounts apply
- Sonnet 5 launch pricing $2/$10 through Aug 2026, then $3/$15
- 1M-token context, no surcharge
- 1M-token context, no surcharge
- Highest capability in the lineup
Use cases
What you can produce with Anthropic API
- Messages API with native multi-turn tool use
- Prompt caching — up to 90% off cached input tokens
- Batch API at 50% off standard rates
- 1M-token context on Opus 4.8 and Sonnet 4.6, no surcharge
- Python and TypeScript SDKs
- Computer Use / agentic tool-calling support
ASEAN Perspective
Anthropic API in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
The Anthropic API remains the most direct route to Claude's reasoning and coding strength, with clean Python and TypeScript SDKs, well-documented tool use, prompt caching that cuts repeated-context cost by 90%, and a 1M-token context window on Opus 4.8 and Sonnet 4.6 at no surcharge. G2 reviewers rate it 4.6 out of 5 across 283 reviews, consistently citing consistency and documentation quality.
It suits teams that want reasoning and coding quality over the cheapest possible tokens, and are willing to build their own infrastructure around a raw API rather than a packaged product. One thing worth flagging: Anthropic's own April 2026 postmortem confirmed six weeks of "Claude feels dumber" complaints came from product-layer changes in Claude Code, not the API or model weights — worth knowing before assuming a quality dip is coming from your own integration. Pricing runs from Haiku at $1/$5 per million tokens up to Opus at $5/$25; batch processing halves both.
What people say
G2 puts Anthropic's product line at 4.6 out of 5 across 283 reviews, with developers consistently pointing to consistency and well-documented endpoints as reasons they stick with the API over switching providers. That reputation took a real hit in March-April 2026, though not from the API itself. Anthropic's own engineering postmortem, published after six weeks of Reddit, X and Hacker News complaints that Claude "felt less capable, more repetitive, and weirdly terse," traced the problem to three unrelated product-layer changes in Claude Code: a March 4 switch of default reasoning effort from high to medium (to fix a UI latency issue), a bug that cleared older thinking context every turn instead of once, and an April 16 verbosity-reduction system prompt that hurt coding quality badly enough to get reverted four days later. All three were resolved by April 20 (Claude Code v2.1.116), Anthropic reset usage limits for affected subscribers, and — the detail that matters for API users specifically — the underlying model weights were never touched. Separately, some Reddit users flagged that Claude Code routes more tasks to the cheaper Haiku model than expected, visible only with verbose logging turned on, which is a Claude Code routing decision rather than an API behavior.
On pricing, Claude Opus 4.8 (released May 28, 2026) runs $5 input / $25 output per million tokens, Claude Sonnet 4.6 sits at $3/$15, and Haiku 4.5 comes in at $1/$5 — with Claude Sonnet 5 available at introductory $2/$10 pricing through August 31, 2026 before reverting to standard rates. Batch processing cuts any of those in half, and prompt caching cuts the cached portion of input by 90%, which matters for anything with a large repeated system prompt or document context. Opus 4.8, Opus 4.7 and Sonnet 4.6 all support a 1M-token context window at flat rates with no premium — a real differentiator against providers that surcharge for long context.
Trustpilot reviews skew toward billing and support-responsiveness complaints rather than model-quality ones — a pattern common to usage-based API products, where the loudest reviews tend to come from account and invoice friction rather than the product itself.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Anthropic API's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Anthropic API unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Anthropic API directly →
Spotted something out of date? Suggest an update →
Anthropic API in the news
More in Code & Dev Tools
