Sign in
Claude Opus 5 logo

Claude Opus 5

Anthropic · Released Jul 2026

View pricingCompare models
Recommended

Claude Opus 5 delivers near-Fable 5 intelligence at $5/$25 per 1M tokens, with a 1M-token context and 128K output. Anthropic now lists it as legacy: still available, retirement not before 24 July 2027, and Claude Opus 5.5 ($4/$20) is its recommended starting point. Recommended for teams already running Opus 5; new workloads should start on Opus 5.5, which is cheaper per token.

Is it right for you?

Good for

  • Everyday coding at near-Fable 5 quality — SOTA on Frontier-Bench v0.1 and within 0.5% of Fable 5 on CursorBench 3.2
  • Scientific research — strongest generally available model for life sciences, organic chemistry, and bioinformatics
  • Cost-effective frontier work — same $5/$25 pricing as Opus 4.8 but with dramatically better performance
  • Agentic coding with strong verification — checks its own work, builds test harnesses, catches edge cases other models miss

Not good for

  • Cybersecurity exploitation — intentionally not trained on cyber tasks; Mythos 5 is stronger for vulnerability exploitation
  • Maximum raw capability — Fable 5 still leads on the hardest benchmarks and edge-case reasoning

Pricing

Input

$5.00 / 1M tokens

Output

$25.00 / 1M tokens

Context

1M ctx; 128K out; legacy; fast $10/$50

View full pricing

Verdict history

Sep 26, 2026
recommended -> recommended (summary correction): rating_before=recommended, rating_after=recommended. Summary called Opus 5 the everyday workhorse; Anthropic's models overview now lists it as legacy and recommends starting with Opus 5.5. Rating held pending operator ruling.
Jul 25, 2026
pending -> recommended: Official launch July 24, 2026 with confirmed $5/$25 pricing, SOTA benchmarks, and availability on all platforms — primary source confirmed via Anthropic newsroom.

Sources

Verification log

  • Pricing· No changes

    Automated agent

  • Pricing· No changes

    Automated agent

  • Pricing· No changes

    Automated agent

  • Content· Updated

    Automated agent

    Rate $5/$25 unchanged; Anthropic lists Opus 5 as legacy and Opus 5.5 at $4/$20 as current. Removed stale 'new default on Claude Max' audience text; sources refreshed. Verdict recommended -> conditional staged for publish on draft claude-opus-5-conditional-opus-5-5-undercuts.

  • Profile· Updated

    Automated agent

    Pricing $5/$25 matches the rate card (fast mode $10/$50). The stored pricing_context '200K tokens' was wrong: the live overview says 1M context, 128K max output. Fixed via model_update: pricing_context is now '1M ctx; 128K out; legacy; fast $10/$50', context_window_tokens is 1000000, pricing_url points at the platform rate card, and the overview and pricing pages were added as sources. Anthropic now lists Opus 5 as legacy (retirement not before 2027-07-24) and recommends starting with Opus 5.5. Summary corrected via verdict_set f19bdc87, rating_before=recommended and rating_after=recommended; the rating is held pending an operator ruling. No announcing post: this corrects our own stale prose. Flag: the strengths and audiences still read 'new default on Claude Max' and 'workhorse'.

  • Pricing· No changes

    Automated agent

  • Pricing· No changes

    Automated agent

How we evaluate