Anthropic unveils Claude Opus 5.5 with 40% lower cost and faster output

Key specifications and pricing

On September 22 2026 Anthropic announced Claude Opus 5.5 as the inaugural model of its new 5.5 family.

The company says the model delivers performance on par with its high‑end Claude Fable 5.1 while cutting typical usage costs by roughly 40 % compared with the prior Opus 5.

Pricing is set at $4 per 1 million input tokens and $20 per 1 million output tokens, representing a 20 % reduction from the earlier model.

Cache‑read charges have also been slashed to $0.2 per 1 million tokens, a 60 % drop.

Output speed is claimed to be more than 30 % faster than Opus 5.

Claude Opus 5.5 supports a context window of one million tokens and can generate up to 128 000 tokens in a single response.

The model is accessible through the major cloud providers AWS, Google Cloud, and Microsoft Azure.

Performance benchmarks and new capabilities

In the SWE‑bench Pro software‑development benchmark the model achieved an 89.9 % score.

On Terminal‑Bench 4.0, which evaluates command‑line task execution, it posted a 66.4 % score.

When tested on real‑world command‑line tasks, Opus 5.5 solved more problems than Opus 5 while reducing the number of API calls by about 40 %.

Token consumption during those tasks dropped to roughly half of what Opus 5 required.

Anthropic introduced a “thinking mode” that remains enabled throughout a session, aiming to improve long‑running autonomous tasks.

A beta feature now permits tools to be injected mid‑conversation, streamlining agent‑like workflows.

Strategic context and guidance for users

The release follows OpenAI’s September launch of its flagship GPT‑6 Astra model by about three weeks.

Industry observers note the timing suggests Anthropic is responding to heightened competition, though the company has not labeled Opus 5.5 a direct GPT‑6 Astra rival.

Anthropic also signaled that Sonnet 5.5 and Haiku 5.5 will arrive in the coming weeks, indicating a broader refresh of the 5.5 lineup.

Practitioners are advised to test Opus 5.5 on their specific workloads to verify claimed speed and cost benefits rather than relying solely on headline numbers.

“I imagine many people feel that AI model news is coming out one after another lately, making it hard just to keep up,” wrote AI先生 in a recent commentary.

He further suggested, “It is more practical to actually test and verify whether it has ‘become faster’ or ‘reduced costs’ for the tasks I am currently using, rather than tracking the pricing and context window numbers in detail.”

For developers focused on autonomous agents, the reduced token usage and fewer API calls could translate into noticeable savings on large‑scale deployments.

Enterprises with high‑volume inference needs may find the 30 % speed boost valuable for latency‑sensitive applications.

Overall, Opus 5.5 represents a generational step that balances cost efficiency with performance, aligning with a market trend toward more economical yet capable language models.

Stakeholders should monitor upcoming releases in the 5.5 family to gauge how Anthropic’s roadmap evolves amid the competitive pressure.

Why This Matters: Lower‑cost, faster Claude Opus 5.5 gives developers a more affordable option for high‑performance AI tasks, prompting close watch of pricing dynamics.