Claude Sonnet 5 shipped alongside the new model generation and took the familiar mid-tier spot: faster and cheaper than the flagship, strong enough for most production workloads.
The price, in numbers
claude-sonnet-5 costs $0.3 per million input tokens and $0.3 per million output tokens. The flagship claude-opus-5 costs $0.4 and $0.4. That's a ten-cent gap per million tokens on each side.
At scale those cents add up, but picking a model on a gap that small is rarely the right call. What matters is whether Sonnet handles your task without Opus.
Where Sonnet 5 matches Opus
- everyday agentic coding: multi-file edits, tests, refactors that follow a clear plan
- API integrations, client code generation from docs
- answering over long context: PR review, ticket summarization, questions about a codebase
- routine agent steps where the task is already well specified
Sonnet 5 holds long context and follows instructions cleanly. That covers most pipelines.
Where Opus earns its keep
The flagship pays off when a task can't be broken into clear steps ahead of time: an unfamiliar architecture that needs to be understood as a whole, a race condition with no working hypothesis, planning under a pile of conflicting constraints. Opus reasons deeper there, and the ten-cent premium stops mattering.
If you're not sure, start with Sonnet. It either solves the task or fails visibly, and switching to Opus after that costs one extra call.
A working example
Anthropic Messages API:
from anthropic import Anthropic
client = Anthropic(
api_key="sk_cdx_your_key_here",
base_url="https://api.claudexia.tech",
)
message = client.messages.create(
model="claude-sonnet-5",
max_tokens=1024,
messages=[{"role": "user", "content": "Split this PR into logical commits."}],
)
print(message.content[0].text)
The same key also works with OpenAI Chat Completions:
from openai import OpenAI
client = OpenAI(api_key="sk_cdx_your_key_here", base_url="https://api.claudexia.tech/v1")
resp = client.chat.completions.create(
model="claude-sonnet-5",
messages=[{"role": "user", "content": "Split this PR into logical commits."}],
)
print(resp.choices[0].message.content)
Both formats run on one endpoint with one key, no need to rebuild infrastructure around a specific SDK.
Sonnet 5 versus Sonnet 4.6
If your prompts are already tuned for 4.6 and things are stable, there's no rush to move: 4.6 stays in the lineup exactly for that reason. For new projects, starting with the 5 makes sense: same price, fresh model, behavior worth testing against your task from day one.
Checking it against your own numbers
Generic advice maps poorly onto a real budget. The site has a calculator: plug in your input and output volumes, compare models on your actual traffic. Per-key stats and budgets, set separately per user and per team, show who's spending what without reconciling a spreadsheet at month end.
In short
Sonnet 5 covers most production traffic at $0.3 per million tokens. Opus earns its place on tasks where the starting point isn't obvious. The price gap between them is small, so decide by the task, not the bill.