Claude API Pricing (October 2026): Which Model Tier to Use, and How to Save

Quick answer: Claude API pricing is per million tokens, billed separately for input and output. Haiku 5.5 is the budget tier, Sonnet 5.5 the mid tier, Opus 5.5 the flagship and Fable 5.1 the premium tier. Batch requests are half price, and cache reads cost a small fraction of normal input. Current rates are on Anthropic’s official pricing page.

Which Claude model tier should you use?

Tier	Best for	Key differences	Official pricing
Claude Fable 5.1	Hardest tasks, top accuracy	Premium tier, the most expensive per token; deepest cache discount	Anthropic pricing
Claude Opus 5.5	Complex coding and agents	Flagship; near-Fable quality at a lower price	Anthropic pricing
Claude Sonnet 5.5 / Sonnet 5	Most production apps	Mid tier; Sonnet 5 keeps its launch price as the standard price	Anthropic pricing
Claude Haiku 5.5	High volume, fast responses	Budget tier; higher rate band for prompts above 100K tokens	Anthropic pricing
Older models (Opus 5, Sonnet 4.6, Haiku 4.5)	Existing integrations	Usually cost more than their 5.5 successors	Anthropic pricing

Why are Opus 5.5 and Haiku 5.5 so much cheaper?

Both launched with price cuts. Anthropic said Opus 5.5 “performs at the level of Claude Fable 5.1 for most tasks, and costs 40% less to run than Opus 5” (@claudeai). It also said Haiku 5.5 “costs around 75% less to run than Claude Haiku 4.5” on average (@claudeai).

Notes from the pricing page:

• Sonnet 5: a price increase that had been scheduled for September 1, 2026 “will not occur”.
• Tokenizer: Claude 4.7 and later models use a newer tokenizer that “produces approximately 30% more tokens for the same text”. Compare costs per task, not just per token.
• Long context: Claude 4.6 and later models (except Haiku 5.5) include the full 1M-token context window at standard pricing.

Which discounts and extra charges apply?

Rule	Effect	Official pricing
Batch API	Half price on input and output	Anthropic pricing
5-minute cache write	1.25x base input	Anthropic pricing
1-hour cache write	2x base input	Anthropic pricing
Cache read	0.1x base input (0.05x on Opus 5.5 and Sonnet 5.5, 0.025x on Fable 5.1)	Anthropic pricing
US-only inference (inference_geo: "us")	1.1x on all token prices, Claude 4.6 and later	Anthropic pricing
Web search tool	Billed per 1,000 searches, plus tokens	Anthropic pricing
Fast mode (Opus 5.5)	Premium rate above standard Opus 5.5	Anthropic pricing

How do I estimate what a request costs?

cost = input_tokens / 1,000,000 × input_rate
     + output_tokens / 1,000,000 × output_rate

Use the current rates from the pricing page. Output costs several times more than input on every Claude tier, and Claude Managed Agents sessions add a session-runtime charge on top of tokens.

How do you get a Claude API key?

1. Create an account in the Claude Console.
2. Add credits or a card on the Billing page (platform.claude.com/settings/billing). New users receive “a small amount of free credits to test the API”.
3. Create a key on the API keys page (platform.claude.com/settings/keys) and keep it server-side.

Full walkthrough: how to get a Claude API key.

Is a Claude subscription cheaper than the API?

Claude.ai subscriptions are billed separately from the API. On claude.com/pricing, Max sits above Pro with higher usage limits, and Max and Team now include monthly API credits at API rates; Pro includes none. For heavy Claude Code use a subscription is usually better value; for apps you need the API. Compare them in Claude Pro vs Max vs API.

How can you cut Claude API costs?

• Default to Sonnet 5.5 or Haiku 5.5, and move up to Opus only for tasks that need it.
• Put long, repeated instructions or documents in a cached prefix.
• Send overnight or bulk jobs through the Batch API.
• Keep Haiku 5.5 prompts under 100K tokens to stay in the lower rate band.
• Try lower effort before a smaller model. An Anthropic engineer noted that Fable 5.1 at low effort often competes with Opus and Sonnet on cost per task while scoring higher (@RLanceMartin).
• Watch your cache hit rate. Fable 5.1 cache reads became 4x cheaper, and the Console can monitor your hit rate (@dani_avila7).
• Audit your code. Anthropic’s /claude-api cost-optimize skill for Claude Code “profiles where your tokens go, ranks levers by savings” and applies the free wins first: caching, input hygiene and batch (@dani_avila7).

Getting errors? See how to fix 401, 429 and 529 errors. Comparing providers? Read OpenAI API pricing and DeepSeek API pricing, or browse the Claude hub.