Anthropic released Claude Opus 4.5 on 24th November 2025—and the numbers are remarkable. The model scored 80.9% on SWE-bench Verified, beating GPT-5.1's 77.9% and Gemini 3's 76.2% to claim the top spot for AI coding. More striking: Opus 4.5 outperformed every human candidate who has ever taken Anthropic's internal engineering examination. Yet despite this premium performance, pricing dropped 66% to just £4 per million input tokens (down from £12) and £20 per million output tokens (down from £60). For UK businesses, this means access to the world's most capable AI reasoning at costs that make practical sense for everyday use.

Last Updated: 25th November 2025 | Reading Time: 8 minutes

What Is Claude Opus 4.5 and When Was It Released?

Claude Opus 4.5 is Anthropic's flagship AI model, released on 24th November 2025 as the most capable system in the Claude 4.5 family. Anthropic describes it as "the best model in the world for coding, agents, and computer use"—a claim supported by its state-of-the-art performance across multiple benchmarks.

The release completes Anthropic's Claude 4.5 lineup following Claude Sonnet 4.5 (September 2025) and Claude Haiku 4.5 (October 2025). This rapid three-model release in two months reflects the intensifying competition between Anthropic, OpenAI, and Google DeepMind.

Claude Opus 4.5 is available immediately via the Anthropic API (model ID: claude-opus-4-5-20251101), AWS Bedrock, Google Cloud Vertex AI, and through Claude.ai for Max, Team, and Enterprise subscribers. GitHub Copilot users can access Opus 4.5 in public preview at a promotional 1x premium request multiplier through 5th December 2025.

How Does Claude Opus 4.5 Compare to GPT-5.1 and Gemini 3?

Claude Opus 4.5 leads all major AI models on SWE-bench Verified with 80.9% accuracy, 3 percentage points ahead of GPT-5.1 Codex Max (77.9%) and 4.7 points ahead of Gemini 3 Pro (76.2%). The benchmark measures real-world software engineering capability across 500 GitHub issues from popular repositories.

Model	SWE-bench Verified	Input (per 1M tokens)	Output (per 1M tokens)
Claude Opus 4.5	80.9%	£4 ($5)	£20 ($25)
GPT-5.1 Codex Max	77.9%	~£2.40 ($3)	~£12 ($15)
Claude Sonnet 4.5	77.2%	£2.40 ($3)	£12 ($15)
Gemini 3 Pro	76.2%	~£2 ($2.50)	~£8 ($10)

Beyond SWE-bench, Claude Opus 4.5 achieved a 10.6% improvement over Sonnet 4.5 on the Aider Polyglot benchmark and leads across 7 of 8 programming languages on SWE-bench Multilingual. On Vending-Bench, which tests multi-step agentic tasks, Opus 4.5 scored 29% higher than Sonnet 4.5.

For UK businesses evaluating AI models, these benchmarks translate to practical advantages: fewer bugs in generated code, more reliable automation workflows, and reduced time spent reviewing AI-generated work.

What Does Claude Opus 4.5 Cost for UK Businesses?

Claude Opus 4.5 costs £4 per million input tokens and £20 per million output tokens at current exchange rates ($5 and $25 respectively). This represents a 66% reduction from Claude Opus 4.1's pricing of £12/$15 input and £60/$75 output—making premium AI reasoning dramatically more accessible.

Pricing Tier	Input (per 1M)	Output (per 1M)
Standard API	£4 ($5)	£20 ($25)
Prompt Caching (write)	£5 ($6.25)	—
Prompt Caching (read)	£0.40 ($0.50)	—
Batch Processing	£2 ($2.50)	£10 ($12.50)

What does this cost in practice? A typical document analysis task (10,000 input tokens, 2,000 output tokens) costs approximately £0.08. A complex coding task with extensive context (100,000 input tokens, 10,000 output tokens) costs roughly £0.60. For comparison, the same tasks on Claude Opus 4.1 would have cost £0.27 and £1.95 respectively.

UK businesses running high-volume AI workloads should explore batch processing, which reduces costs by 50% in exchange for longer processing windows (up to 24 hours). Prompt caching offers another optimisation path—cache reads cost just 10% of standard input pricing.

Did Claude Opus 4.5 Really Beat Human Engineers?

Yes. Anthropic confirmed that Claude Opus 4.5 scored higher than any human candidate in the company's history on their most challenging internal engineering assessment, completing the exam within a 2-hour time limit. This marks the first documented instance of an AI system outperforming human job applicants on a company's own technical hiring test.

The implications for UK businesses are significant. Technical hiring assessments—long considered a reliable filter for engineering capability—may need rethinking when AI systems can outperform human candidates. This doesn't mean AI replaces engineers, but it does suggest AI-assisted coding will become standard practice rather than optional enhancement.

For context, Anthropic's engineering hiring bar is notoriously high. The company employs researchers and engineers responsible for building frontier AI systems. Claude Opus 4.5 exceeding this standard indicates the model can handle complex, multi-step engineering challenges that previously required senior-level human expertise.

UK development teams should consider how this capability changes their workflows. Tasks traditionally requiring senior developer review—architecture decisions, complex debugging, code security analysis—may benefit from Opus 4.5's reasoning capabilities as a complement to human judgement.

What Is the Effort Parameter and How Does It Work?

Claude Opus 4.5 introduces an adjustable "effort" parameter that controls reasoning intensity. At medium effort, the model matches Claude Sonnet 4.5's quality whilst using 76% fewer output tokens—delivering the same results at significantly lower cost and latency.

The effort parameter allows UK businesses to optimise the cost-quality tradeoff for different use cases:

Low effort: Fast responses for straightforward queries, similar to Sonnet performance at reduced cost
Medium effort: Balanced reasoning for most business tasks, matching Sonnet 4.5 quality with fewer tokens
High effort: Maximum reasoning depth for complex analysis, research synthesis, and high-stakes decisions

This flexibility means businesses don't need to choose between cost and capability. Customer service queries can run at low effort, contract analysis at medium effort, and strategic research at high effort—all using the same model with dynamic resource allocation.

For API users, the effort parameter integrates with existing workflows. Teams can set default effort levels per use case and override for specific requests, creating efficient AI operations without sacrificing quality where it matters.

What Are the New Chrome and Excel Features?

Anthropic expanded Claude for Chrome and Claude for Excel alongside the Opus 4.5 release, making these previously pilot-only features more broadly available.

Claude for Chrome is now available to all Max subscribers. The extension allows Claude to interact with web pages directly—reading content, filling forms, navigating sites, and extracting information without manual copy-paste. For UK businesses, this enables automated research workflows, competitor monitoring, and web-based data collection tasks.

Claude for Excel expands to Max, Team, and Enterprise beta users. The integration brings Claude's reasoning capabilities directly into spreadsheet workflows—analysing data patterns, generating formulas, creating visualisations, and automating repetitive calculations. This addresses a significant pain point for UK businesses where Excel remains the dominant tool for financial analysis, reporting, and planning.

Both features leverage Opus 4.5's improved computer use capabilities. Anthropic reports significant improvements in agentic tasks requiring screen interaction, making these integrations more reliable than previous iterations.

For UK SMBs, the Chrome extension reduces time spent on manual web research, whilst the Excel integration accelerates financial analysis and reporting. Neither requires technical setup beyond installing the extension and connecting your Claude account.

What Can UK Businesses Actually Do With Claude Opus 4.5?

Claude Opus 4.5's capabilities translate to practical applications across multiple UK business contexts:

Software Development Teams

With 80.9% accuracy on real-world coding tasks, Opus 4.5 handles complex debugging, code review, and architecture decisions. UK development teams can use it for reviewing pull requests, identifying security vulnerabilities, and generating test cases. The 200,000 token context window accommodates entire codebases for holistic analysis.

Learn more about AI coding tools in our guide to AI coding tools in 2025: what works and what doesn't.

Professional Services

Legal firms can use Opus 4.5 for contract analysis, due diligence research, and regulatory compliance checking. Accounting practices benefit from financial statement analysis and audit preparation support. The model's reasoning capabilities handle nuanced interpretation that simpler AI systems miss.

Research and Analysis

Opus 4.5 excels at synthesising information across large document sets—market research reports, academic papers, competitor analyses. The model identifies patterns and contradictions that human researchers might miss under time pressure, accelerating insight generation.

Strategic Planning

Business leaders can use Opus 4.5 for scenario modelling, risk assessment, and strategic option evaluation. The model's ability to hold complex reasoning chains makes it suitable for multi-factor analysis where simpler AI systems produce superficial outputs.

When to Use Opus 4.5 vs Sonnet 4.5

Not every task requires Opus 4.5's premium capabilities. For UK businesses, the decision framework is straightforward:

Use Opus 4.5 for: complex reasoning, high-stakes analysis, multi-step agent workflows, research synthesis, code architecture decisions
Use Sonnet 4.5 for: everyday coding, content generation, customer support, routine analysis, cost-sensitive applications

See our comparison of Claude Sonnet 4.5 for UK development teams for detailed guidance on choosing between models.

How Do UK SMBs Get Started With Claude Opus 4.5?

UK businesses have several paths to access Claude Opus 4.5, depending on technical requirements and usage volume:

Option 1: Claude.ai Subscription

The simplest approach for most UK businesses. Claude Max subscription (approximately £16/month) provides web access to Opus 4.5, the Chrome extension, and Excel integration. No technical setup required—sign up, subscribe, and start using immediately.

Option 2: Anthropic API

For businesses building AI into applications or running automated workflows. API access requires developer setup but offers full control over prompts, parameters, and integrations. Pay-per-use pricing means costs scale with actual usage rather than fixed monthly fees.

Model ID: claude-opus-4-5-20251101

Option 3: AWS Bedrock

UK businesses with existing AWS infrastructure can access Opus 4.5 through Bedrock. This option supports UK data residency requirements (eu-west-2 London region) and integrates with existing AWS security and billing. Ideal for enterprises with compliance constraints.

Option 4: Google Cloud Vertex AI

Similar to AWS Bedrock, Vertex AI provides managed access to Claude models within Google Cloud infrastructure. Suitable for businesses already using Google Cloud services.

Option 5: GitHub Copilot

Development teams using GitHub Copilot can access Opus 4.5 in public preview. Promotional pricing runs through 5th December 2025 at 1x premium request multiplier, making this an affordable trial option for coding-focused use cases.

Key Takeaways for UK Business Owners

Claude Opus 4.5 represents a significant milestone in AI capability for UK businesses. The model achieves 80.9% on SWE-bench Verified—the highest score of any AI model—whilst pricing dropped 66% to £4/£20 per million tokens. The fact that it outperformed all human candidates on Anthropic's engineering exam signals a new tier of AI capability.

Recommended actions for UK businesses:

Evaluate your current AI usage: Identify tasks where premium reasoning capability would deliver measurable value—complex analysis, high-stakes decisions, research synthesis
Test the effort parameter: Many tasks may achieve sufficient quality at medium effort, significantly reducing costs whilst maintaining Opus-tier capability
Consider the Chrome and Excel integrations: These features can automate manual workflows without requiring API integration or developer resources
Review your model selection: The pricing reduction makes Opus 4.5 viable for use cases previously relegated to Sonnet due to cost constraints

The competitive landscape continues to evolve rapidly. OpenAI released GPT-5.1 on 12th November, Google launched Gemini 3 on 18th November, and now Anthropic completes their Claude 4.5 family. UK businesses have unprecedented choice in AI capabilities—the question is matching the right model to the right use case.

Next Steps for UK Businesses:

Schedule a free AI audit to identify the 3 highest-impact AI opportunities in your operations
Review your GEO audit options to ensure your content appears in AI-powered search results
Book a free 30-minute consultation to discuss how Opus 4.5 could transform your high-value business processes

About the Author: Jake Holmes founded Grow Fast after watching too many businesses waste six figures on AI implementations that looked impressive in demos but failed in production. As a software engineer turned AI consultant, he's spent five years working hands-on with UK development teams and businesses turning over £1-10M, helping them cut through AI hype to find implementations that actually deliver measurable efficiency gains. Book a free 30-minute consultation at grow-fast.co.uk.

Frequently Asked Questions

What is Claude Opus 4.5?

Claude Opus 4.5 is Anthropic's flagship AI model released on 24th November 2025. It achieved 80.9% on SWE-bench Verified (beating GPT-5.1 and Gemini 3) and costs £4 per million input tokens—66% less than its predecessor.

How much does Claude Opus 4.5 cost?

Claude Opus 4.5 costs £4 ($5) per million input tokens and £20 ($25) per million output tokens. Batch processing reduces this to £2/$10, and prompt caching reads cost just £0.40 per million tokens.

Is Claude Opus 4.5 better than GPT-5.1?

For coding tasks, yes. Claude Opus 4.5 scores 80.9% on SWE-bench Verified compared to GPT-5.1 Codex Max's 77.9%. However, GPT-5.1 costs approximately 40% less per token, so the best choice depends on whether accuracy or cost matters more for your use case.

Did Claude Opus 4.5 really beat human engineers?

Yes. Anthropic confirmed that Opus 4.5 scored higher than any human candidate in the company's history on their internal engineering examination, completing it within the 2-hour time limit.

How do I access Claude Opus 4.5 in the UK?

UK businesses can access Opus 4.5 through Claude.ai (Max subscription), the Anthropic API, AWS Bedrock (with eu-west-2 London region for data residency), Google Cloud Vertex AI, or GitHub Copilot (in public preview).

When should I use Opus 4.5 vs Sonnet 4.5?

Use Opus 4.5 for complex reasoning, high-stakes analysis, multi-step agent workflows, and research synthesis. Use Sonnet 4.5 for everyday coding, content generation, customer support, and cost-sensitive applications where Sonnet's 77.2% accuracy is sufficient.

Sources:

Anthropic (24 November 2025): Claude Opus 4.5 Launch Announcement — anthropic.com
TechCrunch (24 November 2025): Anthropic Releases Opus 4.5 — techcrunch.com
GitHub Blog (24 November 2025): Claude Opus 4.5 for GitHub Copilot — github.blog
CNBC (24 November 2025): Anthropic Unveils Claude Opus 4.5 — cnbc.com
VentureBeat (24 November 2025): Claude Opus 4.5 Analysis — venturebeat.com

Claude Opus 4.5: The AI That Beat Human Engineers Now Costs £4 Per Million Tokens