No, Anthropic Has Not Confirmed 'Claude Sonnet 5' or 'Opus 4.8': A Look at an Unverified Claim
Unverified Claim: 'Claude Sonnet 5' and 'Opus 4.8' Cannot Be Confirmed
A claim has been circulating that Anthropic released a model called 'Claude Sonnet 5,' matching the performance of a model called 'Opus 4.8' on agentic benchmarks, at a price of $2 per million tokens. Based on the sourcing available at the time of this research, that claim cannot be verified.
No dated article, official press release, or benchmark report has been located confirming the existence of a model named 'Claude Sonnet 5' or 'Opus 4.8.' The sources reviewed for this piece are general landing and category pages rather than specific evidence tied to this claim.
What the Sourcing Actually Shows
The source set reviewed includes Anthropic's news homepage, its product page, its documentation homepage, Reuters' technology section, TechCrunch's Anthropic tag page, and The Verge's AI section. These are broad, top-level pages rather than individual articles or announcements.
None of these sources contain a title, timestamp, or excerpt referencing 'Claude Sonnet 5,' 'Opus 4.8,' or a comparison between the two on agentic benchmarks. There is no pricing confirmation for a $2 per million token rate tied to either model, and no benchmark data, methodology, or results describing an agentic evaluation that supports the claimed comparison.
Naming Convention Inconsistency
Anthropic's publicly known model lineage includes Claude 3, Claude 3.5, Claude 3.7, and the Claude 4 family, which includes Opus, Sonnet, and Haiku variants. The names 'Sonnet 5' and 'Opus 4.8' do not align with this established pattern.
This inconsistency raises the possibility that the topic reflects speculation, an unannounced internal codename, or a claim that does not correspond to any real Anthropic product. Readers should treat the names themselves as a signal of unverified origin.
Why This Matters for Readers
In a fast-moving AI news cycle, claims about new model releases, pricing, and benchmark performance can spread quickly before they are confirmed. Verifying such claims against primary sources, such as official company announcements or documented benchmark disclosures, remains an important step before treating them as fact.
Amplifying unconfirmed or potentially fabricated model announcements can mislead readers about the actual state of AI product development. Many observers note that skepticism toward specific pricing figures and benchmark comparisons is warranted when no primary documentation is available.
What to Watch For
Readers interested in confirming whether a model along these lines is eventually released should monitor Anthropic's official announcements and changelog directly. Any legitimate release would likely include primary benchmark disclosures, such as agentic task evaluation results, published by Anthropic itself.
Pricing claims, including the $2 per million token figure referenced in this topic, should be cross-checked against Anthropic's official pricing page at the time of any confirmed release, rather than accepted from secondary or unverified sources.