Current Anthropic Claude API Pricing 2026: Claude 5, Opus, Sonnet & Haiku
Current Claude API prices per 1M tokens for Fable 5, Opus 5, Sonnet 5, Opus 4.8, and Haiku 4.5, including prompt caching, Batch discounts, and August 2026 introductory pricing.
Anthropic’s current API lineup is led by Claude Opus 5, Claude Sonnet 5, and the restored Claude Fable 5. Sonnet 5 has introductory pricing through August 31, 2026; its standard rate increases on September 1. Opus 4.8, Sonnet 4.6, and Haiku 4.5 remain available, but they are no longer the whole current lineup.
Claude API Prices Per 1M Tokens
| Model | Input | Cache read | 5m cache write | Output | Context | Status |
|---|---|---|---|---|---|---|
| Claude Fable 5 | $10.00 | $1.00 | $12.50 | $50.00 | 1M | Restored globally July 1 |
| Claude Opus 5 | $5.00 | $0.50 | $6.25 | $25.00 | 1M | Current Opus flagship |
| Claude Sonnet 5 | $2.00 | $0.20 | $2.50 | $10.00 | 1M | Intro price through Aug 31 |
| Claude Opus 4.8 | $5.00 | $0.50 | $6.25 | $25.00 | 1M | Previous Opus generation |
| Claude Sonnet 4.6 | $3.00 | $0.30 | $3.75 | $15.00 | 1M | Previous Sonnet generation |
| Claude Haiku 4.5 | $1.00 | $0.10 | $1.25 | $5.00 | 200K | Current low-cost Claude tier |
Prices are USD per 1 million tokens. The cache-write column uses Anthropic’s 5-minute TTL. A 1-hour write costs 2 times base input: $20/M for Fable 5, $10/M for Opus 5, $4/M for Sonnet 5 during the introductory period, and $2/M for Haiku 4.5.
Sonnet 5 Price Changes on September 1
| Sonnet 5 rate | Through Aug 31 | From Sep 1 |
|---|---|---|
| Input | $2.00 | $3.00 |
| Cache read | $0.20 | $0.30 |
| 5m cache write | $2.50 | $3.75 |
| 1h cache write | $4.00 | $6.00 |
| Output | $10.00 | $15.00 |
Do not use the introductory rate for a September forecast. DevTk.AI’s calculator reflects today’s August price and records the scheduled increase in the model note.
Batch API Prices
Anthropic’s Message Batches API discounts input and output by 50% for asynchronous work.
| Model | Batch input | Batch output |
|---|---|---|
| Fable 5 | $5.00 | $25.00 |
| Opus 5 | $2.50 | $12.50 |
| Sonnet 5, through Aug 31 | $1.00 | $5.00 |
| Haiku 4.5 | $0.50 | $2.50 |
Batch processing and prompt caching can be combined. This is especially useful for evaluations, document pipelines, and agent jobs that reuse large tool schemas or project instructions.
Monthly Cost Example
For 2M uncached input tokens plus 500K output tokens at current standard rates:
| Model | Monthly cost |
|---|---|
| Claude Fable 5 | $45.00 |
| Claude Opus 5 | $22.50 |
| Claude Sonnet 5 | $9.00 |
| Claude Haiku 4.5 | $4.50 |
Sonnet 5 is the main pricing change: at its August rate, this token mix is 60% cheaper than Opus 5 and one third cheaper than the same workload on Sonnet 4.6.
Which Claude Model Should You Use?
- Fable 5: fiction, narrative development, style-sensitive long-form writing, and creative workflows that justify the premium.
- Opus 5: difficult coding, research, long-running agents, and high-impact professional work.
- Sonnet 5: the first model to test for production coding, tool use, assistants, and balanced agent traffic during the introductory period.
- Haiku 4.5: extraction, classification, short answers, routing, and other latency-sensitive lower-cost Claude workloads.
For repeated system prompts and tool definitions, keep prefixes stable and measure cache reads. For non-urgent work, use Batch. For mixed traffic, route routine requests to Haiku or Sonnet and reserve Opus for escalations.
Use the AI Model Pricing Calculator for your token mix, or compare other providers in the AI API Pricing Comparison.
Official Sources
Related Posts
Is Claude Opus 4.8 Worth Upgrading To? Capability, Cost, and the Fable 5 Problem
2026-06-14
GPT-5.6 API Pricing Guide 2026: Sol, Terra & Luna Token Costs
2026-07-14
MCP vs A2A: The Two Protocols Shaping the AI Agent Ecosystem (2026)
2026-03-01