Claude

Anthropic Claude

Anthropic

Anthropic's Claude family leads in coding, reasoning, and writing. Opus 4.7 holds the highest SWE-bench score among commercially available models at 80.8%, with a 1M token context window at standard pricing.

claude.ai
Last updated: April 20, 2026
$5/M
Opus Input
1M
Context Window
128K
Max Output
80.8%
SWE-bench
★★★★★
Coding
★★★★★
Reasoning
★★★★★
Writing
★★★★★
Speed

Available Models

Model Input $/1M Output $/1M Context Max Output Best For
Opus 4.7 (flagship) $5.00 $25.00 1M 128K Complex coding, agent teams, enterprise
Sonnet 4.6 $3.00 $15.00 1M 128K Balanced coding + reasoning
Haiku 4.5 $1.00 $5.00 200K 128K Fast classification, chat, summaries
Opus 4.6 $5.00 $25.00 1M 128K Previous flagship (still available)

Consumer Plans

PlanPriceIncludes
Free$0/moLimited Sonnet 4.6 access
Pro$20/mo5x more usage, Opus 4.7 access
Max$100/mo5x Pro usage, priority access
Max (20x)$200/mo20x Pro usage, highest limits
Team$25-125/seat/moCollaboration, admin controls

Strengths & Weaknesses

Strengths
  • Exceptional coding: 80.8% SWE-bench, best commercially available (Anthropic Docs)
  • Hybrid reasoning with adaptive/extended thinking
  • Constitutional AI for safety and alignment
  • 1M context at standard pricing — no surcharge
  • Industry-leading prose quality for writing tasks
Weaknesses
  • Higher per-token cost than GPT-5.4 and Gemini
  • New tokenizer uses ~1.3x more tokens than before (Finout)
  • Slower latency at full 1M context

Best For

  • Complex software engineering and multi-file code changes
  • Agent teams and agentic workflows
  • Enterprise knowledge work and document analysis
  • Long-context RAG over codebases and legal docs
  • Creative and technical writing

Latest Release NEW

Claude Opus 4.7 — April 16, 2026

  • Strongest coding model commercially available — 80.8% SWE-bench score
  • New tokenizer uses 1–1.35x tokens compared to previous generation
  • Adaptive thinking with max effort parameter for flexible reasoning depth
  • 1M token context window at standard pricing — no surcharge
  • Improved vision capabilities for image and document understanding

Previous Releases

Claude Opus 4.6 — October 2025

  • First model in the Claude 4 family — major leap in coding and reasoning
  • 1M context window, 128K max output tokens
  • Introduced Constitutional AI v2 safety framework
  • Still available via API alongside Opus 4.7

Claude Sonnet 4.6 — October 2025

  • Balanced performance tier — strong coding at lower cost ($3/M input)
  • Same 1M context window as Opus
  • Recommended for most production workloads