Claude Opus 5.5: Anthropic's New Flagship Matches Fable 5.1 at 40% Less Cost — and Lands in GitHub Copilot

What Opus 5.5 actually claims

Anthropic launched Claude Opus 5.5 on September 22, 2026 — the first model in a new Claude 5.5 family and the company's most advanced model to date by its own account — and the headline claim is unusually double-edged for a flagship release. Opus 5.5 performs at the level of Claude Fable 5.1 for most tasks, and it costs about 40% less to run than Opus 5 on typical workloads at default settings. A flagship that matches a cheaper model and undercuts the old flagship on cost is not the standard launch shape, and the unusual part is not the capability claim — it is the cost claim, because most flagship launches are "more capable, more expensive, if you need the ceiling," and Anthropic is saying the opposite on cost while holding the line on capability.

The pricing is where the efficiency claim becomes concrete rather than promotional. Opus 5.5 lists at $4 per million input tokens and $20 per million output tokens on the first-party API, which is 20% cheaper than Opus 5 on input and output, with cached input reads at $0.20 per million — 60% cheaper than the comparable Opus 5 cache rate — and cache writes at $5 for a five-minute window. Batch processing is half price at $2 input and $10 output, and a fast mode is available on the first-party API at $8 input and $40 output. The context window is 1 million tokens with multimodal input, and Anthropic says the model leads on its own benchmarks in agentic coding, computer use, and knowledge work.

The '40% less to run' figure, read carefully

The "40% less to run" figure is the one to read carefully, because it is a typical-workload claim rather than a raw per-token price cut. On raw per-token pricing, Opus 5.5 is 20% cheaper than Opus 5 on input and output, and the larger reduction Anthropic is leading with comes from the combination of cheaper rates, cheaper cache reads, and a model that spends fewer tokens and steps to finish the same work. That is a different kind of efficiency story than "we lowered the list price," and it is the kind of claim that only really lands when developers run the same workload on both models and compare the bills. A 20% per-token price cut is a billing change; a claim that the model finishes the same work in fewer steps and tokens is a workflow change, and the two compound.

Distribution: same-day arrival in GitHub Copilot

Distribution is where the release gets interesting beyond the model and its billing. Anthropic published "Claude Opus 5.5 is available today" at 16:31 UTC on September 22 and put the announcement live at anthropic.com/claude-opus-5-5 the same minute, and the same day GitHub added Claude Opus 5.5 to Copilot for agentic coding, long-running agentic tasks, and knowledge work. GitHub's changelog says that in early testing Opus 5.5 resolved tasks comparably to Claude Opus 5 while using fewer steps and tokens and recovering from errors faster, and the model is rolling out to Pro+, Max, Business, and Enterprise subscribers in a phased rollout with no universal availability date announced yet. That phased rollout means this is a gradual expansion rather than a clean "everyone gets it today" switch, but the direction is clear.

The strategic read inside Anthropic's move is that Opus 5.5 absorbs the job of the old premium tier while Claude Fable 5.1, which shipped earlier in September, stays as the cheaper option for users who do not need the full Opus ceiling. That is a cleaner internal ladder than the one Anthropic had before, and it matches the broader September pattern of frontier labs rebuilding their tiers around "what work actually costs" rather than around raw capability tiers alone. OpenAI did something similar on the same day with GPT-6 Sol and GPT-6 Luna, expanding the GPT-6 family downward, and Anthropic did it by making its top tier cheaper to run than the one it replaces — different shape, same direction.

The GitHub Copilot landing is also a counter-signal to the idea that Anthropic is only available through its own API and Claude apps. Copilot is a major distribution channel, and Opus 5.5 being there the same day it launches means Anthropic is treating developer-tool distribution as a first-class part of the release rather than an afterthought. A model that is only available through its own API is a model developers have to deliberately choose, and a model that is already inside the tool they are using is a model they can try without friction — which is the difference between a model that gets adopted and a model that gets benchmarked once and forgotten.

Which workload is Opus 5.5 for?

If you are choosing between Claude Opus 5.5 and Claude Fable 5.1 for a workload in September 2026, the honest framing is that Opus 5.5 is the model to use when you want the top of Anthropic's capability range and care about the total cost of running the workload rather than the per-token list price. Fable 5.1 remains the cheaper on-paper option and is still the model for users who do not need the Opus ceiling. If you are choosing between Opus 5.5 and GPT-6 Sol for agentic coding or knowledge work, you are choosing between two flagship-adjacent models from labs that both launched on the same day, and the deciding factor is likely to be which ecosystem — Anthropic's API and Copilot, or OpenAI's ChatGPT Work, Codex, and API — your team is already in. Sol is cheaper on raw per-token pricing at $2 in and $10 out against Opus 5.5's $4 in and $20 out, but Anthropic's typical-workload cost claim is the part that could close that gap in practice.

Anthropic's announcement also quietly resets the floor under what a flagship costs to run. A model that performs at the level of the previous premium tier and costs 40% less to run on typical workloads is a model that changes the cost calculus for anyone who was already paying for Opus 5 — not in a dramatic benchmark way, but in a billing way that is more likely to change behavior. That is not a seismic benchmark moment. It is a seismic billing moment, and the billing moments are the ones that actually move teams from "we should evaluate that" to "we turned it on this week."

The phased Copilot rollout is the one thing to watch after the launch itself. Anthropic put Opus 5.5 in Copilot the same day it launched, which is a strong first move, but the absence of a universal availability date means the distribution is gradual rather than immediate. How fast Copilot users actually get access — and whether Anthropic treats Copilot as a permanent distribution channel or a launch-week highlight — will tell you more about the model's reach than the benchmarks will.

— Sources: Anthropic's Claude Opus 5.5 announcement (anthropic.com/claude-opus-5-5, Sep 22 2026); GitHub Copilot changelog (github.blog, Sep 22 2026); 9to5Mac; BNN Bloomberg; Yahoo Finance; LLM Stats; Kingy AI; AI Pricing Guru; Codersera; CellCog; Superpower Daily.