Claude Sonnet 5 is worth using for most everyday work, and if you were still on Sonnet 4.6, you should switch today. For hard analytical or engineering tasks, Opus 4.8 is still the model I reach for. Sonnet 5 sits just below it in capability at a lower price per token.
At TJ Digital, we run most of our daily work in Claude and test every major model as it ships. That efficiency is a big reason we deliver about four times the workload at the same rates.
So which model to reach for on a given task is a call we make every day. Here is my honest read on Sonnet 5, when to use it, and when to skip it.
Is Claude Sonnet 5 an Upgrade Over Sonnet 4.6?
Yes. Sonnet 5 is a blanket improvement over Sonnet 4.6. If you were using 4.6 for anything, move to 5.
Anthropic reports gain across reasoning, tool use, coding, and knowledge work, and the benchmarks back that up. On one agentic coding test, Sonnet 5 scores 63.2%, up from 58.1% on Sonnet 4.6.
It also became the default model on the Free and Pro plans the day it launched. That tells you Anthropic expects almost everyone on the old Sonnet to switch.
How Much Does Claude Sonnet 5 Cost Compared to Opus 4.8?
Sonnet 5 costs about 60% per token of what Opus 4.8 costs. Through August 31, 2026, an introductory rate drops that to roughly 40% of the Opus price. After that, standard pricing takes over.
In dollars, Sonnet 5 runs $3 per million input tokens and $15 per million output tokens at standard rates, with the intro rate at $2 and $10. Opus 4.8 sits at $5 and $25. You can confirm the current numbers in Anthropic’s Sonnet 5 announcement.
Here is the catch. For difficult tasks like software engineering, Sonnet 5 burns through far more tokens than Opus 4.8 does.
It also uses a new tokenizer that turns the same text into roughly 30% more tokens. A lower per-token rate does not always mean a cheaper job. On a hard, multi-step task, the model that finishes in fewer steps can win on total cost.
| Model | Standard Price (input / output per million tokens) | Agentic Coding Benchmark | Knowledge Work | Best For |
| Claude Sonnet 5 | $3 / $15 (intro $2 / $10 through Aug 31, 2026) | 63.2% | Slight edge over Opus 4.8 | Everyday tasks, writing, high volume |
| Claude Opus 4.8 | $5 / $25 | 69.2% | Very strong | Hard analytical and engineering work |
When Should You Use Sonnet 5 Instead of Opus 4.8?
Use Sonnet 5 when you want a capable model for simpler, well-defined work. Think summaries, drafting, data extraction, classification, and everyday questions. It is fast and follows instructions well, and it holds up over a long context.
I have started testing it for writing blog posts, and so far I like it slightly better than Opus 4.8, and it runs a bit faster. That makes it a strong fit for the content workflows we run every day. If you want to see how those come together, I broke down our content machine in another post.
On the knowledge work benchmark, Sonnet 5 actually edges out Opus 4.8 at a lower price. For that whole category of work, it is the better value.
When Not to Use Claude Sonnet 5
I would not hand Sonnet 5 difficult analytical work. If it is the kind of task you would only give to a very smart person, give it to a very smart model.
For complex, multi-step engineering, deep research, or subtle judgment calls, Opus 4.8 plans better and recovers from its own mistakes better. On those jobs, its higher accuracy is worth the higher price. Moving fast only helps if the work is right.
For a fuller map of which model handles which job, I keep a running breakdown of what each model does best. The short version is to match the difficulty of the work to the capability of the model.
Where Do Fable 5 and Opus 5 Fit In?
Two bigger questions sit behind the Sonnet 5 release. The first is Fable 5, Anthropic’s frontier model and, right now, the smartest model available. The open question is whether Anthropic keeps letting us use it on a subscription or forces usage-based pricing.
That matters because subscription rates are heavily subsidized. On a Max plan, you pay only a small fraction, maybe around 4%, of what the same usage would cost through the API.
So that one pricing decision determines whether many of us can afford the most powerful model. I compared Claude Fable 5 against the latest ChatGPT in a separate post.
The second question is Opus 5. It is a little odd that we got Sonnet 5 and Fable 5 before Opus 5, since Opus sits right between those two tiers. I expect that model soon, and I would not be surprised if it triggers a government security review, the way Fable 5 did.
Common Questions About Claude Sonnet 5
Is Claude Sonnet 5 free to use?
Yes. Sonnet 5 is the default model on the Free and Pro plans, and it is available on Max, Team, and Enterprise plans too.
Does Claude Sonnet 5 use more tokens than Sonnet 4.6?
Yes. Its new tokenizer produces roughly 30% more tokens for the same text, so a task can cost more even when the per-token rate looks similar.
Is Sonnet 5 good for writing and content?
Yes. In my testing it writes slightly better than Opus 4.8 for blog posts, and it runs faster, which makes it a good default for content work.
What is the introductory price for Claude Sonnet 5?
Through August 31, 2026, Sonnet 5 runs $2 per million input tokens and $10 per million output tokens. After that it moves to $3 and $15.
Putting the Right AI Model to Work for Your Business
At TJ Digital, we build AI systems that help you show up when your customers search on ChatGPT, Google AI Mode, and Perplexity. We run most of our own work in Claude, and we help small and medium businesses put these tools to work the right way. Get in touch to see what that looks like for your business.