Anthropic Debuts Claude Opus 4.5 — Faster Coding, Smarter Reasoning, Lower Costs
.webp)
Anthropic has rolled out Claude Opus 4.5, the newest
and most advanced model in the Claude 4.5 lineup. Positioned as the company’s
flagship system, this version brings sweeping upgrades in coding performance,
long-context understanding, and multi-step task coordination — all while
consuming significantly fewer tokens than earlier models. Opus 4.5 is now live
through the Claude API and supported cloud partners, stepping in as the
successor to Opus 4.1.
What’s Improved in Claude Opus 4.5
Anthropic’s announcement highlights three core upgrades:
enhanced coding abilities, smarter tool usage, and stronger long-context
reasoning.
1. Big Advances in Coding Performance
Opus 4.5 can now handle complex, multi-stage coding tasks
with far greater efficiency. According to the company, the model solves more
long-range programming problems than its predecessor — while using up to 65%
fewer tokens.
In real-world terms, it can follow extended instructions
without needing oversized prompt windows, thanks to improved planning and more
streamlined internal reasoning.
2. Powerful Multi-Step Task Execution
Anthropic trained the model to handle sophisticated
workflows that involve multiple agents or simultaneous workstreams. During
internal evaluations, Opus 4.5 was capable of:
- Refactoring
two separate codebases at the same time
- Directing
the actions of multiple agents
- Tracking
a high-level plan while executing detailed instructions
A redesigned tool-calling system supports these
capabilities. Instead of loading large tool libraries, the model selectively
pulls in only what is necessary, cutting context usage by as much as 85%.
3. Stronger Long-Context Creativity and Reasoning
For writing and content tasks, Opus 4.5 shows better
consistency across long passages. Anthropic reports that the model can produce 10–15
page chapters while keeping story tone, plot, and character behavior
intact.
It also shows improved performance in 3D spatial
reasoning, delivering clearer descriptions of environments, object
relationships, and complex scenes.
Benchmark Results and Market Position
Anthropic’s internal benchmarks suggest Opus 4.5 has raised
the bar for agentic coding.
On the SWE-Bench Verified test — a key benchmark for autonomous coding
systems — Opus 4.5 scored:
- 80.9%,
outperforming
- Gemini
3 Pro at 76.2%
- GPT-5.1
Codex Max at 77.9%
Lower Costs, Wider Availability
Beyond performance, Anthropic emphasizes reduced operating
costs. For many enterprise tasks, Opus 4.5 delivers equal or better results at nearly
one-third the price of previous models.
The new system is currently accessible:
- Through
the Claude app and website (for paying subscribers)
- Via
the Anthropic API
- On
major cloud platforms like Google Vertex AI and Amazon Bedrock
#ClaudeOpus45 #AnthropicAI #AIUpdate #ArtificialIntelligence
#AIModels #TechNews #CodingAI #LLM #MachineLearning #Innovation #Opus45
