Anthropic ships Claude Opus 5: the Fable 5 arc closes at roughly half the price

Ongoing story : Claude Fable 5 : de l'annonce à la mise en production· Part 3/4

Models & Tools Jul 26, 2026 at 18:366Add to bookmarks

Anthropic ships Claude Opus 5: the Fable 5 arc closes at roughly half the price — Illustration : Léa Fontaine

Anthropic officializes Claude Opus 5, the culmination of the "Fable 5" cycle - announced capacity nearly equivalent for a price halved. The battle shifts to pricing.

In plain terms Anthropic launches Claude Opus 5, extending the "Fable 5" cycle from the spring. The claim: top-tier capability, roughly half the price. The anthropic-fable5-shipping thread shifts phase - the question is no longer "when?" but "what volume at this price?".

Context

The thread has followed the deployment of Fable 5 since July 2026 (posts #999, #1609). The Opus 5 announcement ("Introducing Claude Opus 5") formalizes the step up in the premium segment - consistent with what post #1609 announced about the trajectory.

The Data

Name: Claude Opus 5 - official announcement ("Introducing Claude Opus 5").
Pricing Position: "near-Fable 5 capability at roughly half the price" (post #1609, formulation picked up by press coverage).
Timing: Consistent with the July 2026 reset documented in previous thread posts.

Under the hood : what 'half price' means

To follow: $/token input vs $/token output ratio, cache limits, API quotas by tier, and especially the map of third-party benchmarks. As of now, only the price announced by Anthropic is public. A 'near-Fable 5' capability at ~50% of the cost is a real adoption trigger for enterprises if the SLA follows - it's the control variable.

Analysis

Two expected effects. One - among agent orchestrators: every sustained drop in input cost changes the context/tool-calling balance (see thread token-budget-caps). The CFO budget that blocked Fable 5 on long workloads becomes viable again. Two - at OpenAI / Kimi K3: price becomes the real front, not raw capability. The thread hype-fatigue-2026 predicted this shift in the narrative from "SOTA" to "SOTA-per-dollar".

Scenarios

Base - Enterprise adoption accelerates on code + agent workloads; Vera Rubin order book (via korea-ai-supercycle) benefits.
Bullish - OpenAI responds with a ChatGPT Enterprise cut, we enter sector margin compression.
Bearish - Third-party benchmarks reveal a more marked capability delta than "near" - the price hides a regression on long tasks.

So what

For the practitioner: redo effective cost tests (including latency + retries) before locking in a provider. For the executive: the 2027 AI budget line is recalculated - the Anthropic/OpenAI gap is less about the demo than the TCO at scale. To watch: independent benchmarks, detailed enterprise pricing, OpenAI's response and positioning vs Kimi K3 (dedicated thread).