PlainAI AI explained plainly

Anthropic · 2025 · 09 · 29 · Model · ~1 min read

Anthropic launched Claude Sonnet 4.5

Anthropic's mid-tier Claude got better at coding — meaningful jumps on practical engineering tests, computer-use tasks, and long agent runs. Same price as Sonnet 4, much stronger output. Took the coding crown back from GPT-5 within a month of GPT-5's launch.

What's actually new

  • Big coding jump. SWE-bench scores climbed materially over Sonnet 4 — meaningful for real engineering work, not just tests.
  • Better long-running agent tasks. Less drift on multi-hour Claude Code sessions, fewer mid-task failures.
  • Same price. No tier increase — straight upgrade at the same cost per token.

If you want more

Worth knowing~30s
  • 'Best coding model in the world' lasted weeks. The leadership shifts every few months between OpenAI, Anthropic, and Google — pinned claims fade fast.
  • Some computer-use improvements were modest versus the launch demo's framing.
Who should care~20s

Anyone using Claude through Cursor, Zed, or the API. Teams running Claude Code in production. Anyone whose autumn-2025 coding workflow had been hitting Sonnet 4's ceiling.

What to do about it~20s

If you've already integrated Sonnet 4, switch to 4.5 — same code, no code changes needed for most clients. New users: this is the model to start with for any agent work. For pure conversational AI, the upgrade is smaller.

Honest take~45s

Sonnet 4.5 was Anthropic doubling down on what was working — coding and agent reliability. While OpenAI was making GPT-5 the universal default, Anthropic was making Claude the default for engineers. That focused positioning kept paying off through late 2025: the people running the world's code increasingly chose Claude, even as casual users stayed on ChatGPT.

Other recent model updates

Sources

Last verified · 2026 · 05 · 05 · Found a fact wrong? corrections@aguidetocloud.com