Anthropic’s Claude Sonnet 5, released June 30, 2026, is doing something unusual for a frontier-adjacent model: competing on price. Introductory API pricing of $2 per million input tokens and $10 per million output tokens runs through August 31, 2026 — after which it rises to $3/$15. That is a 50% increase on the exact same workload, and it turns the next few weeks into a genuine decision window for anyone running AI at volume.
Key Facts
- Released June 30, 2026 — Anthropic’s most agentic Sonnet-class model, successor to Sonnet 4.6
- Intro pricing $2 / $10 per million tokens through August 31, then $3 / $15
- 1M-token context window with context compaction, up to 128K output per response
- Now the default model on Claude Free and Pro plans; available on Max, Team, and Enterprise
- Headline benchmark: 85.2% on SWE-Bench Verified — approaching Opus 4.8 territory at a fraction of the price
The Pricing Math, Worked Out
Percentages hide the real impact, so here is a concrete example. Take a mid-size production workload: 100M input tokens + 20M output tokens per month — roughly what a busy AI feature or a small agent fleet consumes.
Same Workload, Two Prices
| Until Aug 31 | From Sep 1 | Change | |
|---|---|---|---|
| Input (per 1M tokens) | $2 | $3 | +50% |
| Output (per 1M tokens) | $10 | $15 | +50% |
| Example: 100M in + 20M out / month | $400/mo | $600/mo | +$200/mo |
Run your own numbers in our free AI API cost calculator — it covers 29 models side by side.
Two honest framings of the same numbers: the intro window is a 33% discount on what Anthropic considers the model’s real price — and September 1 brings a 50% increase on whatever you built during the summer. Standard $3/$15 simply returns Sonnet to its classic price point; the discount, not the increase, is the anomaly.
What You Get for the Money
The benchmark story is why this pricing matters. Sonnet 5 posts 85.2% on SWE-Bench Verified, 63.2% on the harder SWE-Bench Pro, 80.4% on Terminal-Bench 2.1, and 81.2% on OSWorld-Verified — numbers that meaningfully narrow the gap to Anthropic’s own flagship Opus 4.8 on coding, tool use, and computer use. Add the 1M-token context window with compaction and 128K output ceiling, and this is a near-flagship agent engine at a mid-tier sticker.
Convly’s Take
Our price-performance index has tracked one pattern all year: the model market is splitting into a barbell — frontier flagships on one end, ruthless efficiency models on the other, and a dying middle. Sonnet 5 is the most serious attempt yet to rescue the middle: near-flagship agentic capability priced closer to the efficiency tier, at least until August 31. The intro window is also a classic anchor strategy — teams that wire Sonnet 5 into production this summer are unlikely to re-platform over a 50% bump in September. Decide with the $3/$15 price in your spreadsheet, not the $2/$10 one.
How to Use the Window
- Evaluate now, commit at standard price. Run your pilots while tokens are 33% off — but green-light only what still pencils out at $3/$15.
- Front-load heavy one-off jobs. Backfills, large-corpus summarization, bulk classification, dataset generation — anything batchable is effectively on sale until August 31.
- Right-size per task. A 1M-token context is powerful and expensive by nature: long contexts multiply input billing. For routine steps, smaller models remain dramatically cheaper — the 100×+ capability-per-dollar spread across the market hasn’t gone anywhere.
Run the Numbers Yourself
- AI API Cost Calculator — compare 29 models on your real token volumes
- AI Price-Performance Index 2026 — our original study: the 114× spread in cost per unit of capability
- Live AI Models Database — specs, pricing, and hardware requirements in one table
FAQ
How much does Claude Sonnet 5 cost?
$2 per million input tokens and $10 per million output tokens until August 31, 2026. From September 1, standard pricing is $3/$15 — a 50% increase.
Is Claude Sonnet 5 better than Opus 4.8?
No — Opus 4.8 remains Anthropic’s flagship. But Sonnet 5 narrows the gap substantially on coding and agentic benchmarks (85.2% SWE-Bench Verified) while costing far less, which makes it the better value for most production workloads.
What is the context window of Claude Sonnet 5?
1 million tokens, with context compaction and up to 128K output tokens per response.
Is Claude Sonnet 5 free to use?
Yes — it is now the default model on Claude’s Free and Pro plans. The $2/$10 pricing applies to API usage.
