On June 30, 2026, Anthropic officially introduced Claude Sonnet 5, the newest addition to its Sonnet line of AI models within the broader Claude family. The new model is positioned as the direct successor to Claude Sonnet 4.6, and Anthropic describes it as the most agentic Sonnet released so far, meaning it is significantly better at planning multi-step tasks, using real-world tools such as browsers and terminals, and carrying complex jobs through to completion with minimal human intervention.
For years, Sonnet-class models, starting with versions 3.5, 3.6, and 3.7, were the ones that kicked off the agentic AI era thanks to their strong coding and tool-use skills. Over time, however, the most noticeable gains in this area shifted toward the larger, more expensive Opus-class models. Sonnet 5 narrows that gap in a meaningful way: Anthropic states that its performance now sits close to that of Claude Opus 4.8 on many agentic tasks, while remaining considerably cheaper to run, making it a practical option for both individual developers and large organizations.
Key Updates and Features in Claude Sonnet 5
Sonnet 5 ships with a range of substantial improvements over its predecessor:
- Stronger reasoning and more effective tool use throughout multi-step tasks.
- Improved coding ability, including sustained software engineering work across messy, real-world codebases without stopping short.
- Better results on agentic search and computer-use benchmarks, two key metrics Anthropic uses to measure a model’s ability to work independently.
- An adjustable effort level, letting users balance cost and performance depending on the complexity of each task.
- An updated tokenizer, similar to the change introduced with Opus 4.7, which can affect how many tokens the same input text produces.
- Safety improvements, with internal evaluations showing lower rates of undesirable behavior compared with Sonnet 4.6.
Sonnet 5 vs. Sonnet 4.6: The Main Differences
To understand how much progress this release represents, it helps to compare it directly with the previous version, Claude Sonnet 4.6:
- Agentic performance: Sonnet 5 clearly outperforms Sonnet 4.6 on planning and autonomous execution, and on some tasks it approaches Opus 4.8-level capability.
- Cost efficiency: thanks to adjustable effort levels, Sonnet 5 offers a much wider range of cost-to-performance options than its predecessor.
- Safety and reliability: Anthropic’s evaluations show that Sonnet 5 has lower rates of hallucination and sycophancy than Sonnet 4.6, along with a better ability to refuse harmful requests and resist prompt injection attempts.
- Cyber capabilities: despite overall improvement, Sonnet 5 remains far less capable than Opus-class models at executing hacking and exploit-development tasks, which Anthropic views as a positive from a safety standpoint.
Availability and Pricing
As of launch, Claude Sonnet 5 is the default model on both the Free and Pro plans, and it is also available to Max, Team, and Enterprise users, as well as inside Claude Code and the Claude Platform via the API.
On pricing, Anthropic launched the model at an introductory rate of 2 dollars per million input tokens and 10 dollars per million output tokens, valid through August 31, 2026. After that date, standard pricing takes effect at 3 dollars per million input tokens and 15 dollars per million output tokens. Anthropic has also raised rate limits across Chat, Cowork, Claude Code, and the Claude Platform to accommodate higher-effort usage.
Safety and Responsible Use
Anthropic paid particular attention to safety in this release. According to the official System Card, Sonnet 5 showed an overall improvement in behavior compared with its predecessor, with a lower rate of undesirable behaviors across the company’s broad automated behavioral audits.
Default cyber safeguards, similar to those applied in Opus 4.7 and 4.8, are also enabled to detect and block potentially dangerous usage in real time. Readers interested in deeper technical detail can review the full System Card directly on Anthropic’s official website.
Why Claude Sonnet 5 Matters for Users and Businesses in Saudi Arabia
With Saudi Arabia’s digital transformation plans accelerating under Vision 2030, and artificial intelligence increasingly entering local business sectors, the launch of advanced yet competitively priced models like Claude Sonnet 5 represents a meaningful opportunity for regional startups and developers to build agentic AI solutions without taking on the heavier usage costs typically tied to larger models.
Whether in customer service automation, data analysis, or software development, this model can offer strong performance at a lower operating cost, aligning with the broader push toward adopting AI technology more economically across both the public and private sectors.
Frequently Asked Questions About Claude Sonnet 5
What is Claude Sonnet 5?
It is the newest AI model in Anthropic’s Sonnet family, designed to be the most capable at independent, agentic work among Sonnet releases, through planning tasks, using tools, and executing multiple steps without constant human intervention.
When was Claude Sonnet 5 released?
Anthropic officially announced Claude Sonnet 5 on June 30, 2026.
What is the difference between Claude Sonnet 5 and Claude Sonnet 4.6?
Sonnet 5 outperforms its predecessor in reasoning, tool use, coding, and overall agentic performance, alongside noticeable safety improvements and reduced rates of hallucination and sycophancy.
Is Claude Sonnet 5 better than Claude Opus 4.8?
Not overall, but it comes close on a number of specific agentic tasks, delivering comparable performance at a much lower cost.
How much does Claude Sonnet 5 cost through the API?
The introductory price through August 31, 2026 is 2 dollars per million input tokens and 10 dollars per million output tokens, moving afterward to 3 dollars per million input tokens and 15
