.png){kind=small}
On November 24, Anthropic's engineering team introduced Claude Opus 4.5 with a 67% price reduction and significant performance improvements.
The model achieved 80.9% accuracy on SWE-bench Verified, becoming the first to break the 80% barrier on this challenging software engineering benchmark.
This positions Anthropic to re-enter the top tier of coding-focused frontier models after their mid-tier Sonnet model had been outperforming the older Opus 4.1, giving users little reason to pay for the premium tier.
How It Performs
The model demonstrated what developers describe as "calm intelligence", producing working solutions on the first attempt rather than requiring iterative fixes.
In real-world testing:
- Outperformed all human candidates on Anthropic's notoriously difficult two-hour engineering exam
- Maintained coherent reasoning across complex, multi-step problems
- Led across 7 out of 8 programming languages on the SWE-bench Multilingual
- Showed creative problem-solving that occasionally outsmarted benchmark expectations
As Anthropic said, developers comparing Opus 4.5 with GPT-5.1 and Gemini 3 report differing strengths depending on task type.
Opus 4.5 maintained clarity across long reasoning chains while remembering previous decisions, though Gemini 3 Pro dominated visual UI tasks and GPT-5.1 brought faster responses for simpler queries.
New Features
Anthropic introduced an "effort parameter" giving developers control over computational intensity:
Medium effort: Matches Sonnet 4.5 performance while using 76% fewer output tokens
High effort: Exceeds Sonnet 4.5 by 4.3 percentage points while using 48% fewer tokens
The developer platform also gained:
- Enhanced plan mode for Claude Code with upfront clarifying questions
- Desktop app support for running multiple local and remote sessions in parallel
- Unlimited conversation length through automatic context summarization
- Improved context management and memory capabilities for agentic tasks
Pricing
API pricing dropped to $5 per million input tokens and $25 per million output tokens—down from $15/$75. This makes Opus 4.5 significantly more affordable than its predecessor while maintaining premium positioning against:
- GPT-5.1: $1.25/$10 per million tokens
- Gemini 3 Pro: $2/$12 per million tokens
Availability
Claude Opus 4.5 is available through:
- Claude web, mobile, and desktop applications
- API access using model ID claude-opus-4-5-20251101
- AWS, Google Cloud, and Microsoft Azure platforms
- GitHub Copilot integration
Additional features now open to subscribers:
- Chrome extension (all Max users at $100/month+)
- Excel automation (Max, Team, and Enterprise tiers)
- Removed Opus-specific usage caps
- Increased overall limits for Max and Team Premium users
Takeaway
The three-way competition between Anthropic, Google, and OpenAI has clarified each model's strengths.
Developers report using Opus 4.5 as their default for architectural planning and complex reasoning, Gemini 3 for visual UI work, and GPT-5.1 for conversational speed.
The shift from cost-per-token to cost-per-solved-task economics suggests enterprises will increasingly choose models based on reliability over raw speed, with Opus 4.5 positioning itself as the precision tool for critical software engineering workflows.
.png)
.svg)
.png)
.png)
