How much does the Anthropic Claude API cost?

Claude API prices per million tokens are: Claude Fable 5.1 (and its restricted-access counterpart Mythos 5.1) at $10 input / $50 output; Claude Opus 5.5 (released September 22, 2026) at $4 input / $20 output; Claude Sonnet 5.5 (released September 28, 2026) at $2 input / $10 output; and Claude Haiku 4.5 at $1 input / $5 output. Earlier models remain available: Opus 5 and Opus 4.8/4.7/4.6 at $5 / $25, Sonnet 5 at $2 / $10, and Sonnet 4.6 at $3 / $15. Prompt caching reads cost 10% of the input rate on most models (Fable 5.1's cache reads are a flat $0.25 per million), and the Batch API gives a 50% discount on both input and output.

What does Claude Fable 5.1 cost on the API?

Claude Fable 5.1, released September 1, 2026, costs $10 per million input tokens and $50 per million output tokens — the same rate as its restricted-access counterpart Mythos 5.1. Cache reads are a flat $0.25 per million tokens, a 75% reduction from Fable 5's $1.00 cache-read rate.

How much does prompt caching save on the Claude API?

A cache read costs 10% of the base input price (0.1x). Writing to the cache costs 1.25x the base input price for the 5-minute cache or 2x for the 1-hour cache. Caching pays off after a single read for the 5-minute duration, so for repeated system prompts or long documents it sharply reduces input cost.

Does the Claude API have a free tier?

There is no ongoing free tier. New users receive a small amount of free credits to test the API, after which usage is billed per token at the rates above. For free Claude access, use the Claude consumer app's free tier instead of the API.

Anthropic API pricing (2026): the Claude per-token cost table

Q: What is the cheapest Claude API model?

Claude Haiku 4.5 is the cheapest current Claude model at $1 per million input tokens and $5 per million output tokens. You can cut that further: a prompt-cache read costs $0.10 per million tokens (10% of input), and the Batch API halves both input and output to $0.50 / $2.50 per million. For even lower cost on simple tasks, Google's Gemini Flash tier undercuts Haiku.

Short answer: Per million tokens, the Claude API costs $10 in / $50 out for Fable 5.1 (the top model, released Sept 1, 2026), $4 in / $20 out for Opus 5.5 (released Sept 22, 2026), $2 in / $10 out for Sonnet 5.5 (released Sept 28, 2026), and $1 in / $5 out for Haiku 4.5 — the cheapest current model. A prompt-cache read costs 10% of the input rate (Fable 5.1's cache reads are a flat $0.25/MTok), and the Batch API takes 50% off both input and output. Verified against Anthropic's official pricing page.

Claude API model pricing (per 1M tokens)

Model	Input	Output	Best for
Claude Fable 5.1 (top model, Sept 2026)	$10	$50	Flagship frontier model; Mythos 5.1 is the same model under restricted-access safeguards
Claude Opus 5.5 (Sept 22, 2026)	$4	$20	Hardest reasoning, agentic and critical work at below-Fable cost
Claude Sonnet 5.5 (Sept 28, 2026)	$2	$10	Production default; best quality-to-cost balance
Claude Opus 5 / 4.8 / 4.7 / 4.6	$5	$25	Prior Opus releases, still available
Claude Sonnet 5	$2	$10	Prior Sonnet release, still available
Claude Sonnet 4.6	$3	$15	Older Sonnet; now costs more than Sonnet 5.5
Claude Haiku 4.5 (cheapest)	$1	$5	Routing, classification, high-volume simple tasks

Prices are USD per million tokens (MTok) for the first-party Claude API with default global routing. Fable 5.1, Opus 5.5, Sonnet 5.5, Opus 5, Opus 4.8/4.7/4.6 and Sonnet 4.6 include the full 1M-token context window at standard per-token pricing (Fable 5.1, Opus 5.5 and Sonnet 5.5 have a 128K-token max output). Haiku 4.5 has a 200K context window. Older models (Opus 4.1 and earlier, at the legacy $15/$75 Opus rate) are deprecated.

What is the cheapest Claude API model?

Haiku 4.5, at $1 / $5 per million tokens, is the cheapest current Claude model. You can push the effective cost lower in three ways, which stack:

Prompt caching: a cache read is 0.1x input — $0.10 per million tokens on Haiku.
Batch API: 50% off both directions — $0.50 in / $2.50 out per million.
Right-sizing: route easy requests to Haiku and reserve Opus for genuinely hard reasoning.

If you need to go cheaper than Haiku for trivial tasks, Google's Gemini Flash tier undercuts it — see OpenAI & Claude API alternatives.

Prompt caching and Batch API discounts

Feature	Multiplier vs base input	Effect
5-minute cache write	1.25x	Pays off after one cache read
1-hour cache write	2x	Pays off after two cache reads
Cache read (hit)	0.1x	90% off repeated context
Batch API	0.5x in & out	50% off async (non-time-sensitive) jobs

Web search as a server tool is billed separately at $10 per 1,000 searches plus token costs. Caching and batching discounts stack with each other. Claude Fable 5.1 is the exception on caching: instead of the 0.1x multiplier, its cache reads are priced at a flat $0.25 per million tokens — a 75% cut from Fable 5's $1.00 cache-read rate.

Worked example: cost of a typical Sonnet 5.5 request

A request that sends 20,000 input tokens and generates 2,000 output tokens on Sonnet 5.5:

Input: 20,000 × $2 / 1,000,000 = $0.040
Output: 2,000 × $10 / 1,000,000 = $0.020
Total: $0.060 per request (the same request on the older Sonnet 4.6, at $3 / $15, costs $0.090)

If 16,000 of the input tokens are served from a prompt cache (0.1x), the input drops to (4,000 × $2 + 16,000 × $0.20) / 1,000,000 = $0.0112, taking the request to about $0.031 — roughly half.

How Claude API pricing compares to OpenAI

At the top tier, Claude Fable 5.1 ($10 in / $50 out) is priced identically to OpenAI's GPT-6 Astra (released Sept 2026). In the middle, Claude Sonnet 5.5 and OpenAI's GPT-6.1 Sol both cost $2 / $10, with Claude Opus 5.5 ($4 / $20) sitting between the tiers. OpenAI keeps the edge at the cheap end — GPT-6 Luna ($0.10 / $0.50) undercuts Haiku 4.5 ($1 / $5). For the full side-by-side, see ChatGPT vs Claude API cost and the OpenAI API pricing table.