Anthropic Opus 5.5 arrives with 40% cost reduction, delivering Fable 5.1‑level performance at lower token prices, 60% cheaper cache reads, and 30% faster output than Opus 5.

Anthropic Opus 5.5 arrives with 40% cost reduction

Anthropic Opus 5.5 arrives with 40% cost reduction, delivering performance comparable to Claude Fable 5.1 on most work while costing significantly less to run than Opus 5. Launched on September 22, 2026, Opus 5.5 is the first model in Anthropic’s new Claude 5.5 family and is positioned as a high‑efficiency flagship for coding, agentic workflows, and professional knowledge tasks.

Anthropic says the model not only costs less per token but also uses fewer tokens per task, resulting in an overall ~40% drop in typical workload costs compared with Opus 5. Output generation is also more than 30% faster, making Opus 5.5 attractive for interactive coding, research, and agent loops where latency and cost both matter.

Performance: Fable 5.1‑level at lower cost

Opus 5.5 is designed to bring frontier‑class capability down the price ladder:

• On Anthropic’s internal evaluations, it performs at the level of Claude Fable 5.1 on most professional, coding, and knowledge work.
• It leads Anthropic’s reported benchmarks for agentic coding, computer use, and knowledge tasks, often outperforming other models even when they run at their highest effort settings.
• The model is described as a “hybrid reasoning” system built for serious coding and AI agents, with a 1M‑token context window.

This places Opus 5.5 as a more affordable alternative to Fable 5.1 for users who want near‑top performance without paying flagship prices for every task.

Pricing and cost savings

Anthropic highlights two layers of savings:

1. Lower token prices vs Opus 5
2. Fewer tokens per task due to improved efficiency

Token pricing (per 1 million tokens)

Cache reads are especially important for agentic and coding workloads, where large contexts are reused across many steps. Anthropic says cache reads make up the majority of those costs, so the 60% cut from $0.50 to $0.20 has a big impact on total spend.

Overall cost impact

Combining:

• 20% lower input/output prices,
• 60% lower cache‑read prices, and
• improved token efficiency (fewer tokens per task),

Anthropic estimates that typical token‑billed workloads cost about 40% less on Opus 5.5 than on Opus 5 at default settings.

Speed improvements

In addition to cost, Opus 5.5 is optimised for latency:

• Output generation is more than 30% faster than Opus 5.
• A Fast mode is available in Claude Code and the Claude Platform, offering up to 2.5× speed for scenarios where latency is critical.
• Fast mode pricing: $8 per million input tokens and $40 per million output tokens.

This makes Opus 5.5 suitable both for cost‑sensitive batch workloads and for interactive tasks like live coding assistance and real‑time agent coordination.

Availability

Opus 5.5 is available immediately across Anthropic’s ecosystem:

• Claude API – as claude-opus-5.5 for developers and enterprises.
• Claude Code – for AI‑assisted coding workflows.
• Claude apps (web, desktop, mobile) – accessible to Pro, Max, Team, and seat‑based Enterprise users, with usage limits (e.g., five‑hour windows on higher tiers).
• Cloud marketplaces – on Amazon Web Services, Google Cloud, and Microsoft Azure.

Regular ChatGPT‑style users do not need to call the API directly to benefit; they can select Opus 5.5 in supported Claude interfaces where available.

Strategic context

The launch comes shortly after:

• OpenAI introduced GPT‑6 Sol and Luna, cutting prices by 50% vs GPT‑5.6 Sol/Luna.
• Intense competition on both capability and cost among frontier AI providers.

By releasing a model that matches Fable 5.1 on most work but costs ~40% less to run, Anthropic is pushing high‑end capability down the price curve while maintaining its focus on safety, alignment, and enterprise‑grade reliability.

Summary: Anthropic Opus 5.5 arrives with 40% cost reduction, offering Fable 5.1‑level performance, 20% lower token prices, 60% cheaper cache reads, and 30% faster output than Opus 5. It is available now via the Claude API, Claude Code, Claude apps, and major cloud marketplaces.