Claude Opus 5.5: The Top Model Just Got Cheaper To Run
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Claude Opus 5.5: The Top Model Just Got Cheaper To Run on ThorstenMeyerAI.com

Prime Big Deal Days · Oct 6–7Offer from Amazon

Get the latest gadgets delivered free — and shop member deals

  • Fast, free delivery on millions of items
  • Access to Prime Big Deal Days deals on October 6–7
  • Prime Video, Amazon Music and more included
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

Anthropic has launched Claude Opus 5.5, a new AI model that is 20% cheaper to operate and over 30% faster than its predecessor. The update emphasizes cost efficiency, speed, and improved safety features, positioning it as a leading choice for enterprise AI workloads.

Anthropic has introduced Claude Opus 5.5, its latest flagship AI model, which now costs approximately 20% less to operate than the previous Opus 4 model. The release also features a more efficient architecture, enabling the model to generate outputs more than 30% faster, and includes new subscription benefits such as higher usage limits and flexible rate resets. This development marks a significant step in the competitive AI landscape, as Anthropic aims to deliver high-performance models at reduced operational costs. For more context, see the rationale behind favoring Claude Opus 5.5 for AI benchmarks.

Claude Opus 5.5 is described by Anthropic as performing at the level of Claude Fable 5.1 on most tasks, with a maximum Intelligence Index score of 58, the highest measured so far. The model’s costs per 1 million tokens are set at $4 for input and $20 for output, representing a 20% reduction compared to previous versions. Notably, the costs associated with cache reads have dropped by 60%, which is significant because cache reads constitute the majority of costs in agentic and coding tasks, leading to an overall 95% discount on cached input processing. Additionally, the model now produces output more than 30% faster, with an optional ‘Fast’ mode available at up to 2.5 times the speed for $8 per 1 million tokens.

Anthropic claims that the cost savings are due both to lower per-token prices and reduced token consumption per task, though independent testing by Artificial Analysis suggests that at maximum effort, token usage remains comparable or slightly higher than previous models. The company emphasizes that typical workloads under default settings benefit most from the cost reductions. Early user feedback indicates that Opus 5.5 outperforms its predecessor in real-world tasks such as code reviews, bug detection, and codebase analysis, often completing tasks in less time and with fewer tokens used. To learn more about the model’s capabilities, visit our analysis of ByteDance’s AI model. The model’s improved safety and clarity in communication are also highlighted, with reports of fewer hallucinations and more reliable outputs.

At a glance
announcementWhen: announced March 2024
The developmentAnthropic announced the release of Claude Opus 5.5, highlighting significant cost reductions and performance improvements over previous models.

Claude Opus 5.5 at a glance

Anthropic’s September 22, 2026 flagship leads the independent Intelligence Index, cuts token prices, and makes the effort setting the biggest lever on your bill.

58Artificial Analysis Intelligence Index at max effort, the highest measured
−60%Cache read price, the main cost of agentic and coding work
30%+Faster output than Opus 5, per Anthropic

New prices

Per 1M tokensOpus 5Opus 5.5Change
Input$5.00$4.00−20%
Output$25.00$20.00−20%
Cache reads$0.50$0.20−60%
Cache writes$6.25$5.00−20%

Fast mode, up to 2.5× speed, costs $8 input and $40 output per 1M tokens.

The effort dial is the real cost lever

Intelligence Index score (in the bar) and cost per index task (above it), by effort level.

$0.55
42
$1.34
51
$1.82
54
$3.46
56
$5.98
58
low
medium (default)
high
xhigh
max

Medium gets 51 of 58 points for about a fifth of the max-effort cost. Four of the five levels sit on the intelligence-versus-cost frontier.

“40% cheaper” depends on the setting

−40%

Anthropic: cost versus Opus 5 at default settings on typical workloads, from lower prices and fewer tokens per task.

≈ level

Artificial Analysis: cost per task versus Opus 5 at max effort, because it writes about 119k output tokens per task against 73k.

Both are true. Turn the dial up and you pay for the extra thinking. Early testers report low or medium effort now matches Opus 5 at high.

Where it leads, and where it doesn’t

Leads (independent testing)

  • AA‑Briefcase: 1822 Elo, +143 over Fable 5.1
  • GDPval‑AA: 1846 Elo across 44 occupations
  • Humanity’s Last Exam: 61.4%
  • SciCode: 66.9%
  • Terminal‑Bench 4.0: 59.6%, level with GPT‑6 Astra

Still trails

  • CritPt (physics reasoning)
  • AA‑LCR (long‑context reasoning)
  • GDP.pdf (professional documents)

Anthropic itself says benchmark margins are now a less reliable guide to real‑world differences.

Safety and safeguards

Better

  • Best score yet on a ~2,000‑scenario behavioral audit
  • About 85% fewer attempts to cross containment boundaries than Opus 5
  • Tied for lowest prompt‑injection success rate in Gray Swan’s test
  • Zero data retention available; EU AI Act watermarking

Plan around

  • Most cybersecurity tasks re‑route to Opus 4.8
  • Biology safeguards match Fable 5.1; verification programs available
  • Thinking mode can no longer be switched off
  • Anthropic reports it often suspects it’s being evaluated

What to do this week

Lower your effort setting first. It’s likely a bigger saving than the price cut.
Budget in cost per task, not cost per token. Only your own workload settles it.
Running agents unattended? The safety results matter more than two index points.
In security or life sciences? Test the safeguard path before you migrate.
ThorstenMeyerAI.comSources: Anthropic (pricing, vendor benchmarks, safety) and Artificial Analysis (independent evaluation and per‑effort model pages). Figures as of 23 September 2026.

Enhanced Cost-Effectiveness and Performance Impact

The release of Claude Opus 5.5 signifies a major shift in AI operational economics, as it offers a more affordable option for enterprise users without sacrificing performance. The significant reduction in cache read costs and faster output generation make it an attractive choice for tasks requiring repetitive processing or code reruns, which are common in software development, automation, and knowledge work. This development could influence how organizations allocate AI resources, potentially lowering barriers to adoption and enabling more complex workflows within existing budgets.

Furthermore, the improvements in safety and output clarity address common concerns about hallucinations and unreliable communication, making the model more suitable for client-facing work and critical decision-making. As the AI market becomes increasingly competitive, Anthropic’s focus on cost efficiency combined with high performance positions Claude Opus 5.5 as a strategic player, possibly shifting industry standards around operational costs and model capabilities.

Amazon

AI developer workstation

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Recent AI Model Developments and Competitive Landscape

Earlier this week, OpenAI announced GPT‑6 Sol and Luna, with prices cut in half, signaling a move toward more affordable AI solutions. In response, Anthropic launched Claude Opus 5.5, which not only matches or exceeds the performance of previous models but also emphasizes cost reductions. Historically, AI companies have competed on raw capability, but recent releases highlight a shift toward balancing performance with operational efficiency. Anthropic’s approach of lowering costs while maintaining high performance reflects an industry trend driven by enterprise demand for scalable, cost-effective AI tools.

Prior to this, models like Opus 5 had established a reputation for strong performance in code and knowledge work, but at higher operational costs. The new release aims to redefine the economics of AI deployment, especially for tasks involving repetitive or high-volume processing, by drastically reducing cache read costs and increasing speed. This strategic move comes amid broader industry efforts to make AI more accessible and sustainable at scale.

“At its lowest effort setting, Opus 5.5 detects a higher percentage of bugs in code reviews than earlier models, demonstrating its practical efficiency.”

— Deloitte AI team

Amazon

high performance coding laptop

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Unconfirmed Aspects and Ongoing Evaluations

While early testing indicates significant cost and speed improvements, some claims—such as token usage reduction at default settings—are based on company assertions and may vary in real-world deployment. Independent benchmarks at maximum effort show similar or slightly higher token consumption, suggesting that actual savings depend on workload and effort levels. Additionally, the long-term safety and reliability of the model in diverse applications remain to be fully validated through broader industry testing.

Amazon

AI model training hardware

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Next Steps for Adoption and Industry Impact

Organizations interested in Claude Opus 5.5 are expected to evaluate its performance across various use cases, including coding, knowledge work, and client-facing deliverables. As more users adopt the model, further independent testing will clarify its real-world cost savings and safety features. Industry analysts anticipate that this release will accelerate competitive dynamics, prompting other AI providers to prioritize operational efficiency alongside raw capability. Anthropic may also release additional updates or new models to expand on these advances in the coming months.

Amazon

enterprise AI server

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Key Questions

How much cheaper is Claude Opus 5.5 compared to previous models?

According to Anthropic, Opus 5.5 costs about 20% less per 1 million tokens for input and output, with cache read costs dropping by 60%, representing a significant reduction in operational expenses.

What performance improvements does Opus 5.5 offer?

Opus 5.5 generates output more than 30% faster than Opus 5 and scores higher on several intelligence benchmarks, especially excelling in knowledge work and coding tasks.

Are there any safety or reliability benefits with Opus 5.5?

Yes, early tests report fewer hallucinations and more reliable, clearer communication, making it suitable for client-facing and critical applications.

Will the cost savings impact AI deployment strategies?

Yes, the reduced operational costs could enable organizations to scale AI use more broadly, especially in repetitive or high-volume tasks, potentially reshaping deployment strategies.

What is still uncertain about Opus 5.5’s performance?

While early results are promising, some claims about token efficiency are based on company assertions, and independent testing at maximum effort shows comparable or higher token use, so real-world savings may vary depending on workload and effort levels.

Source: ThorstenMeyerAI.com

FALL

Fall Picks

As an affiliate, we earn on qualifying purchases.

You May Also Like

The Future Of Work Is AI: How RingCentral Leads The Way

OpenAI reports RingCentral is integrating AI across engineering and operations, signaling a shift to AI-native workplace practices, though details remain unclear.

Experts Warn AI Could Pose A 10%+ Risk Of Human Extinction Following Researcher Departure

A researcher from Anthropic warns there is a more than 10% chance AI could cause human extinction, raising urgent safety concerns.

The Odyssey’s Hades Scene Was Done With Practical Effects

Confirmed: ‘The Odyssey’ employed practical effects for its Hades scene, highlighting a shift from CGI reliance. Details on the techniques and impact remain ongoing.

Porting My 1993 Amiga Game To Godot, With An LLM Reading The 68000 Assembly

A developer successfully ported a 1993 Amiga game to Godot with an LLM interpreting 68000 assembly, completing the process in a single evening.