Claude Sonnet 5
The next-gen Sonnet. Priced to replace GPT for daily work.
Claude Sonnet 5 is the current mid-tier Claude model, and for most work it is the one to reach for first. It carries the same 1M-token context window as Opus 5 while costing a fraction as much per token, which makes it the default rather than the compromise.
Anthropic rates its comparative latency Fast, against Moderate for Opus. It is built for work you run constantly — production endpoints, RAG pipelines, batch jobs, test generation — where the bill is the sum of thousands of calls rather than a handful of hard problems.
We sell it at half Anthropic’s list price. Same model id, same weights, same API shape. The discount is on the invoice, not on the model.
What it costs you per month
Real-world budget scenarios. Numbers are simple sums — official list price vs llmrelay's 50% off tier.
Anthropic’s $2.00 / $10.00 list price for Sonnet 5 is introductory and runs through 31 August 2026. Standard pricing of $3.00 / $15.00 takes effect 1 September 2026. Our price stays at half list either way, so budget $1.50 / $7.50 here from September. Worth knowing now if you are modelling costs for a quarter rather than a week.
$0.0047 for a full code review
Rate cards quote price per million tokens, which is not a unit anyone budgets in. So we ran one concrete task and read the billed amount out of our own logs: 592 output tokens, 20.4s, $0.0047.
The same task on Claude Opus 5 cost $0.0152 — this model is 3.27× cheaper per task.
One real code review: find a concurrency bug in a Python billing function, identify two lesser issues, and rewrite it correctly. temperature=0, identical prompt to both models. Measured 31 July 2026. The figure is what our billing system actually charged, not an estimate from the rate card. One run, one task — a cost measurement, not a quality benchmark. Full method and both answers side by side →
What it's good at
- +Priced under GPT-5
- +Sub-second latency
- +Strong at structured output
Best for
- — Production APIs
- — RAG
- — Batch inference
Tools that ship well with Claude Sonnet 5
Pick Claude Sonnet 5 when
- +You are starting a project and have not measured which tier you actually need. Start here and escalate the specific tasks where you can see it failing — that order is cheaper than starting on Opus and trying to trim later.
- +The workload loops: CI review comments, batch classification, structured extraction, test generation. At high call volume the per-call gap against Opus becomes the entire budget.
- +You need long context on a budget. The 1M window is the same as Opus 5, so putting a large codebase or document set in front of the model costs less than half as much.
- +You are the review layer yourself. If you read and verify the output anyway, you are paying twice for a second opinion you do not use.
Choose something else when
- !Work where breadth of review matters more than the bill — auth logic, migrations, anything touching money. On our own measured code review Opus 5 found the same core issues but also flagged smaller ones Sonnet 5 passed over.
- !Long agent chains where one bad step poisons everything downstream. The per-task premium on Opus 5 is small next to the cost of a restart.
- !Creative and narrative work, where Fable 5 is the purpose-built model in this family.
Questions people ask about Claude Sonnet 5
How much does Claude Sonnet 5 cost?
Anthropic’s list price is $2.00 per million input tokens and $10.00 per million output tokens. On llmrelay it is $1.00 and $5.00 — exactly half list. That introductory list price runs through 31 August 2026, after which standard pricing of $3.00 / $15.00 applies and our half-list price becomes $1.50 / $7.50. Billing is prepaid per token from one credit pool, with no subscription.
Is Sonnet 5 good enough for coding, or do I need Opus 5?
For most coding work Sonnet 5 is enough. We sent one identical code review to both models on 31 July 2026: both found the concurrency bug, the SQL injection and the unsafe cache, and Sonnet 5’s fix was arguably the tighter of the two — a single atomic conditional UPDATE that pushes correctness into the database. It cost $0.0047 against Opus 5’s $0.0152. What the extra money bought was breadth, not a better answer: Opus 5 additionally caught an unused import and added input validation. Start on Sonnet 5 and escalate where you see it fall short.
What is the context window on Claude Sonnet 5?
1,000,000 tokens, with up to 128,000 output tokens. That is the same context window as Claude Opus 5, so long-context work does not require paying for the flagship tier.
Is Sonnet 5 actually faster than Opus 5?
Anthropic rates Sonnet’s comparative latency Fast against Moderate for Opus, so it should be. Our own measurement did not show it: 20.4 seconds against 19.7 for the same 2000-token-capped request. One run each is not a timing benchmark, so read that as no meaningful difference observed on a short request rather than evidence against the rating.
Can I use Sonnet 5 with Claude Code, Cursor or Cline?
Yes. We serve an OpenAI-compatible and Anthropic-compatible endpoint, so any tool that lets you set a base URL and an API key works. Point it at https://api.llmrelay.dev/v1 and use the model id claude-sonnet-5 — nothing else in your setup changes.
Try Claude Sonnet 5 at half the price
Free to create an account, no subscription, $10 minimum top-up — enough to run this model against a real task and judge quality yourself.
Get API key →