Claude 5 models

9 versions from Anthropic, reviewed by the TopReviewed AI panel.

ModelScoreStatusReleasedPrice in/outContext
Claude Fable 5
hardest long-horizon work money can buy
9.5GA2026-06-08$10 / $501MReview →
Claude Opus 4.8
production agentic coding workhorse
9.2GA2026-05-27$5 / $251MReview →
Claude Opus 4.7
agentic coding at the frontier
9.0GA2026-04-15$5 / $251MReview →
Claude Sonnet 4.6
best value production workhorse
8.8GA2026-02-16$3 / $151MReview →
Claude Opus 4.6
prior frontier, stable production target
8.6GA2026-02-04$5 / $251MReview →
Claude Opus 4.5
the Opus price reset, stable target
8.0GA2025-11-23$5 / $25200KReview →
Claude Haiku 4.5
high-volume low-latency worker
8.4GA2025-10-14$1 / $5200KReview →
Claude Sonnet 4.5
stable legacy workhorse
7.5GA2025-09-28$3 / $15200KReview →
Claude Opus 4.1
legacy snapshot, maintenance only
5.5GA2025-08-04$15 / $75200KReview →

Strongest: Claude Fable 5

The first publicly available Mythos-class model — a tier Anthropic explicitly positions above Opus. Fable 5 is the same underlying model as the restricted Claude Mythos 5, wrapped in AI-classifier safeguards that route sensitive sessions (cybersecurity, biology/chemistry, distillation attempts) to Opus 4.8. The numbers are a step change: SWE-bench Verified 95.0% and SWE-bench Pro 80.3% versus 88.6/69.2 for Opus 4.8, with state-of-the-art vision and million-token focus. Priced at $10/$50 per Mtok — double Opus — and included free on paid Claude plans until June 22, 2026.

Full review →

Best value: Claude Haiku 4.5

Claude Haiku 4.5 is Anthropic's fastest, lowest-cost model, released October 15, 2025, delivering near-frontier coding and agent performance at $1/$5 per 1M tokens. The headline: it matches or exceeds Sonnet 4 on coding and computer use at roughly one-third the cost and more than twice the speed, and it is the first Haiku with extended thinking. For a buyer, the single sentence is this: route the long tail of work here, and only pay Sonnet or Opus when a job actually needs them. - Provider: Anthropic - Released: 2025-10-15 - Status: GA (current Haiku tier) - Context window: 200,000 tokens - Max output: 64,000 tokens - Modalities: text, image - Knowledge cutoff: February 2025 reliable (training cutoff July 2025) - Headline price: $1 input / $5 output per 1M tokens

Full review →