Curiosity · AI Model

Claude Sonnet 4.5

Claude Sonnet 4.5 is Anthropic's September 2025 mid-tier model — the one that crossed 70% on SWE-bench Verified and briefly held the best-coding-model title. It introduced the Claude 4 agent stack (long tool loops, context editing, memory tools) and remained the default Sonnet tier until Sonnet 4.6 replaced it in early 2026.

Model specs

Vendor
Anthropic
Family
Claude 4
Released
2025-09
Context window
200,000 tokens
Modalities
text, vision, code
Input price
$3/M tok
Output price
$15/M tok
Pricing as of
2026-04-20

Strengths

  • First Sonnet to cross 70% on SWE-bench Verified at launch
  • Stable across very long agent trajectories (many tool calls in a row)
  • Strong computer-use numbers — good for browser automation
  • Well-tuned prompt caching behaviour for long-context apps

Limitations

  • Context window capped at 200K vs 1M in Sonnet 4.6
  • Superseded on most benchmarks by Sonnet 4.6 and Opus 4.7
  • Extended-thinking output tokens priced the same as normal outputs, inflating cost

Use cases

  • Coding agents running multi-hour autonomous sessions
  • Computer-use workflows via Anthropic's computer-use tool
  • Tool-heavy support and ops agents
  • RAG pipelines at Claude-4 quality with moderate cost

Benchmarks

BenchmarkScoreAs of
SWE-bench Verified≈77%2025-09
OSWorld (computer use)≈62%2025-09
MMLU-Pro≈82%2025-09

Frequently asked questions

What is Claude Sonnet 4.5?

Claude Sonnet 4.5 is Anthropic's September 2025 mid-tier model in the Claude 4 family. At launch it achieved best-in-class scores on coding benchmarks like SWE-bench Verified and introduced agent-friendly features such as context editing and the memory tool.

Should I use Sonnet 4.5 or Sonnet 4.6?

For new projects in 2026, use Sonnet 4.6 — it has a longer context window and slightly better benchmarks. Sonnet 4.5 remains a safe choice for existing pipelines tuned to its behaviour.

What is the context window of Claude Sonnet 4.5?

Claude Sonnet 4.5 supports a 200,000-token context window. For very long documents, Sonnet 4.6 or Opus 4.7 with 1M context are a better fit.

Does Claude Sonnet 4.5 support computer use?

Yes. Sonnet 4.5 scored around 62% on OSWorld at launch, making it the first Sonnet strong enough to recommend for production browser-automation agents using Anthropic's computer-use tool.

Sources

  1. Anthropic — Claude Sonnet 4.5 — accessed 2026-04-20
  2. Anthropic — Pricing — accessed 2026-04-20