Editorial illustration for Anthropic's Claude Opus 5 Cuts Token Use 26%, Matches Top-Tier AI Performance
Claude Opus 5 Cuts Costs 26%, Matches Top AI Performance
Anthropic put out Claude Opus 5 on Friday, pricing it at $5 per million input tokens and $25 per million output tokens, the same rate as the Opus 4.8 model it replaces. The company says it delivers close to the intelligence of Claude Fable 5, its most capable model, at roughly half the cost. Opus 5 is live now across Anthropic's platforms, takes over as the default model on the Claude Max subscription tier, and stands as the strongest option available to Claude Pro subscribers.
The launch is less about topping benchmark charts than about where AI spending actually happens. Anthropic isn't calling Opus 5 its smartest system. That title stays with Fable 5, and competitors still beat Anthropic in specific tasks.
The bet here is different: most paid AI work, the company argues, falls into a middle tier of difficulty where a model that's fast and cheap but still close to frontier-level beats one that's marginally smarter but far more expensive to run. That's the logic Anthropic is using to carve up its own lineup into distinct jobs, a structure a company spokesperson laid out to VentureBeat.
Anthropic is not claiming Opus 5 is its smartest model — that distinction still belongs to Fable 5, and rival systems retain an edge in certain domains. Instead, the company is making a subtler argument that may matter more to enterprise buyers: that the most economically important AI work happens in a middle band of difficulty, where near-frontier intelligence delivered efficiently and cheaply beats frontier intelligence delivered expensively.
Why this matters
For developers and founders building on Claude, the token-efficiency number matters more than the headline price tag. Anthropic held pricing flat at $5/$25 per million tokens between Opus 4.8 and Opus 5, so any real savings come from Harvey's reported 26% cut in output tokens at comparable reasoning quality, not from a discount. That's a meaningful shift for anyone running high-volume agentic workloads, where token count, not sticker price, drives the bill.
Fundamental Research Lab's early read on hard financial-modeling tasks suggests the gains may hold up outside legal use cases, but two customer quotes are a thin sample size for a company claiming near-parity with its flagship model. We'd want to see broader benchmarks before treating this as settled. Still, the framing here is telling: Anthropic isn't selling more intelligence, it's selling the same intelligence for less compute.
If that trend continues across releases, the real competitive battleground for enterprise AI buyers becomes cost-per-task efficiency, not leaderboard scores. Worth watching whether rivals respond on efficiency rather than raw capability next.
Related Reading
Browse all Anthropic news →
Business & Startups Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy
Business & Startups Kling launches Video O1, all-in-one model with MVL bridge using transformer
Business & Startups DeepSeek Seeks More Capital Weeks After USD 7B Funding Round
Related Topic Trump cracks down on Anthropic after Amazon tip; staff largely foreign
Open Source
Related Topic Anthropic says Claude writes >90% of its code; AI pause button urged
LLMs & Generative AI
From Archives Microsoft launches cost-cutting AI models in shift from single flagship approach
From Archives Runway launches AI model router based on its creative team's evaluation expertise
From Archives Anthropic expands voice mode to Gmail, Slack apps
LLMs & Generative AI
From Archives Anthropic Beta Tests Claude Security Plugin for Terminal Vulnerability Scanning
Open Source
Common Questions Answered
How much does Claude Opus 5 reduce token usage compared to its predecessor?
Claude Opus 5 cuts output token usage by 26% compared to Opus 4.8 while maintaining comparable reasoning quality. This efficiency gain is particularly significant for developers running high-volume agentic workloads, where token consumption directly impacts operational costs.
Why did Anthropic keep Claude Opus 5's pricing the same as Opus 4.8?
Anthropic maintained the same pricing of $5 per million input tokens and $25 per million output tokens between Opus 4.8 and Opus 5. The real cost savings for users come from the 26% reduction in output tokens required, rather than from a price discount.
How does Claude Opus 5 compare to Claude Fable 5 in terms of intelligence and cost?
Claude Opus 5 delivers close to the intelligence of Claude Fable 5, Anthropic's most capable model, at roughly half the cost. However, Fable 5 remains Anthropic's smartest model, and the company positions Opus 5 as the better choice for enterprise work that doesn't require frontier-level intelligence.
What is Anthropic's strategic argument for Claude Opus 5 over frontier models?
Anthropic argues that the most economically important AI work happens in a middle band of difficulty, where near-frontier intelligence delivered efficiently and cheaply outperforms frontier intelligence delivered expensively. This positioning suggests that cost-effective, token-efficient models like Opus 5 better serve enterprise buyers than premium frontier models.
Which Claude subscription tier now uses Claude Opus 5 as its default model?
Claude Opus 5 is now the default model for the Claude Max subscription tier and stands as the strongest option available to Claude Pro subscribers. The model is live across all of Anthropic's platforms as of its Friday launch.
Further Reading
- Introducing Claude Sonnet 5 - Anthropic
- Pricing - Claude Platform Docs - Anthropic Docs
- Claude Opus 4.5: Token Efficiency Finally Makes Opus Viable - Adam Holter
- Anthropic rolls out Opus 4.5 as token efficiency and conversation memory improve - Mugglehead
- Claude API Pricing 2026: Opus 4.8, Sonnet 4.6, Haiku 4.5 ... - Metacto
