Opus5.app Starter Pack — $29

The comparison · July 25, 2026

Opus 5 vs Fable 5: when “nearly as smart, half the price” is enough.

Anthropic's own positioning makes this comparison unusually honest: Fable 5 is smarter, Opus 5 is cheaper and easier to access. The practical question is where the gap actually bites.

Opus 5Fable 5
API price (per 1M in/out)$5 / $25$10 / $50
Claude ProIncluded — strongest model on ProUsage credits only (one-time $100 included)
Claude Max / Team PremiumIncluded — the default modelIncluded up to 50% of weekly limits
Capability"Nearly all" of Fable 5's intelligence (Anthropic-reported)The top of Anthropic's published line
Safety classifiers~85% fewer interventions (Anthropic-reported)Strictest tier; 30-day retention policy applies
Data retentionExempt from the 30-day policy30-day retention policy

The honest decision rules

Default to Opus 5 if you're on Pro (it's the best you get without credits), you run high-volume agentic work on the API (half rate compounds fast), or your work keeps tripping Fable 5's stricter classifiers.

Stay with Fable 5 if you're on Max (it's already included up to the 50% cap — use it for planning, judging, and the genuinely hard steps) or your task provably needs the extra capability — the classic pattern is Fable 5 plans and reviews while a cheaper model executes.

Both at once is the real answer for most serious setups: flagship judgment where it matters, cheap tokens for volume. That split-stack arithmetic — with real dollar numbers per workflow — is what the Field Guide does weekly; read a full free issue to see the format.

Independent benchmark rounds for Opus 5 were still landing at the time of writing; we'll update this page as third-party numbers (not just Anthropic's) come in. Leave your email below to get that update.

Free briefing

Opus 5 changes weekly — get one email when something real moves.

Pricing shifts, plan changes, honest benchmarks. No spam; from the team behind Fable5.app.