Overview
- Tool name: Claude Models (Fable 5 / Opus 4.8 / Sonnet 5 / Haiku 4.5)
- Developer: Anthropic
- Official website: https://www.anthropic.com/api
- Documentation: https://platform.claude.com/docs
- Category: AI Models
Claude is Anthropic's model family and a consensus top tier for coding and agentic work. The current lineup (four tiers):
- Claude Fable 5: the next-generation flagship for long-running autonomous agents, at $10 input / $50 output - the most expensive tier on the market and the ceiling for the hardest tasks
- Claude Opus 4.8: the workhorse for complex agentic coding and enterprise workloads, $5/$25; a fast mode delivers up to 2.5x speed at 2x pricing
- Claude Sonnet 5: high-performance coding and agent balance, standard $3/$15 (an introductory $2/$10 ran through August 31, 2026)
- Claude Haiku 4.5: fastest and cheapest at $1/$5, for high-throughput simple tasks
API pricing (official, per million tokens, October 2026)
| Model | Input | Output | Cache write (5min) | Cache read |
|---|---|---|---|---|
| Fable 5 | $10 | $50 | $12.50 | $1.00 |
| Opus 4.8 | $5 | $25 | $6.25 | $0.50 |
| Sonnet 5 | $3 | $15 | $3.75 | $0.30 |
| Haiku 4.5 | $1 | $5 | $1.25 | $0.10 |
- Batch processing is 50% off across the lineup; US-only inference at 1.1x
- Pricing verified against anthropic.com/api and the official pricing PDF (effective 2026-06-29)
Ecosystem and capabilities
- Coding agents: Claude Code (the official CLI) plus deep integrations with Cursor, Cline, JetBrains and GitHub Copilot - the de facto standard inside IDEs
- Built-in tools: web search and fetch, code execution, the MCP connector, Files API, structured outputs, memory, and context editing that automatically clears less-relevant tool calls near token limits
- Managed agents: composable APIs for building and deploying agents at scale
Who it is for
- The hardest tasks: Fable 5 targets long-horizon autonomous agents (engineering jobs that run for hours or days)
- Coding and IDE work: Opus 4.8 / Sonnet 5 power Cursor, Claude Code and friends
- Cost-sensitive agent loops: Sonnet 5 with caching (reads at $0.30) and 50%-off batch
- High-throughput light tasks: Haiku 4.5
Limitations
- The flagship's $50 output is the priciest in the industry; model your usage carefully
- Context windows vary by SKU (see the official model pages); verify before long-context work
- No open weights, no self-hosting
Frequently asked questions
What is the most capable Claude model?
Fable 5 ($10/$50), positioned as the next-generation flagship for long-running agents; for everyday coding, Opus 4.8 or Sonnet 5 are the mainstays.
What does the Claude API cost?
Four tiers: Fable 5 $10/$50, Opus 4.8 $5/$25, Sonnet 5 $3/$15, Haiku 4.5 $1/$5 per million tokens (official anthropic.com pricing). Batch is half price across the board.
Claude vs GPT-6 vs Gemini 3?
Choose Claude for top coding and agent quality; Gemini 3 Pro ($2/$12) or DeepSeek V4 for the cheapest frontier tier; GPT-6 / Gemini for ecosystem and multimodal breadth.
Is there a free tier?
Claude.ai has a free tier; the API is usage-based with no permanent free allowance (new accounts may get trial credits).
*This page is not affiliated with Anthropic. Pricing verified as of October 2026 (anthropic.com/api); confirm current figures on the official site.*
