Claude Opus 5 Arrives in Claude Code
Claude CodeAnthropic released Claude Opus 5, positioning it as the new default model on Claude Max and the strongest option on Claude Pro, available immediately in Claude Code. Opus 5 keeps Opus 4.8's pricing ($5/$25 per million input/output tokens) while closing much of the performance gap with the pricier Fable 5 tier, particularly on coding, debugging, and long-running agentic tasks. Claude Code's /model picker now highlights Opus 5 as the new default, fast mode has moved off Opus 4.7 onto Opus 5 and Opus 4.8, and the bundled claude-api skill now defaults to Opus 5 with a documented migration path from 4.8.
Key Takeaways
- Opus 5 keeps Opus 4.8's price tag ($5/$25 per Mtok) while closing much of the gap with the pricier Fable 5 tier on coding and agentic benchmarks.
- It's the new default on Claude Max and the top available model on Claude Pro, shipping directly into Claude Code as of v2.1.219.
- Fast mode moved off Opus 4.7 entirely: /fast now only applies to Opus 5 and Opus 4.8.
- Subagent nesting depth tripled, from 1 to 3 by default, enabling deeper agent-delegation chains.
- Scientific and computer-use benchmarks improved notably, with Anthropic citing a 10.2 point jump on internal chemistry evaluations over Opus 4.8.
- Early reviews are mixed on tone but positive on value, praising raw capability while flagging more conservative guardrail behavior.
Sources & Mentions
4 external resources covering this update
Anthropic launches Claude Opus 5, a cheaper AI model for coding, agents and enterprise workflows
VentureBeat
Anthropic's Claude Opus 5 AI model rivals Fable 5 and is cheaper
CNBC
Claude Opus 5 review: this model is brilliant (but annoying)
Lenny's Newsletter
Anthropic debuts Opus 5 model as company preps for IPO later this year
Yahoo Finance
A New Default Opus Tier
Anthropic shipped Claude Opus 5 on July 24, 2026, and it landed in Claude Code the same day via CLI version 2.1.219. The model becomes the default on Claude Max and is now the strongest model available to Claude Pro subscribers. Pricing holds steady at $5 per million input tokens and $25 per million output tokens, unchanged from Opus 4.8, with fast mode running roughly 2.5x faster at 2x the base price.
Anthropic's own benchmarking claims Opus 5 approaches Fable 5-level intelligence at half the cost. On Frontier-Bench, the model reportedly doubles Opus 4.8's score. On OSWorld 2.0, a computer-use benchmark, Opus 5 is said to match Fable 5's peak score using roughly a third of the compute budget, and it posts a meaningfully higher score on Anthropic's internal organic chemistry and protein-sequence evaluations. Third-party coverage from VentureBeat and CNBC frames the release primarily as a cost story: a model priced at the Opus tier that behaves close to the flagship tier.
What Changed Inside Claude Code
Beyond making Opus 5 selectable, this release adjusts several defaults around it:
- The
/modelpicker now highlights only the newest model's name, so the visual emphasis follows each release rather than staying pinned to "Opus." - Fast mode dropped Opus 4.7 entirely;
/fastnow applies only to Opus 5 and Opus 4.8. - The bundled
claude-apiskill defaults to Opus 5, with in-tool migration guidance for teams still targeting Opus 4.8. - Subagents can now nest up to depth 3 by default (previously capped at depth 1), a change that ships alongside Opus 5 to support deeper agentic delegation chains.
Reception
Early third-party reaction is mixed but largely favorable on cost-to-capability grounds. Lenny's Newsletter's review called the model "brilliant (but annoying)," citing strong raw capability paired with more conservative guardrail behavior. Reporting from VentureBeat and CNBC both emphasize that Opus 5 undercuts Fable 5's price while closing much of the intelligence gap, and note the launch lands as Anthropic prepares for a widely reported IPO later this year.