Claude Sonnet 5 release: what Anthropic shipped on June 30
Claude Sonnet 5 came out on June 30, 2026, and Anthropic says it performs close to Opus 4.8 at lower prices. The announcement from that day presents Sonnet 5 as a step up from Sonnet 4.6. The details below describe the model as it was at launch.
Claude Sonnet 5 is billed as the most agentic Sonnet
Anthropic calls Claude Sonnet 5 the most agentic Sonnet model it has released. An agentic model works through a multi-step task on its own: it picks the next action rather than answering one prompt and stopping. According to Anthropic, Sonnet 5 can plan, use tools such as browsers and terminals, and run autonomously at a level that needed larger, more expensive models only a few months earlier.
"Claude Sonnet 5 is built to be the most agentic Sonnet model yet." ／ "It can make plans, use tools like browsers and terminals, and run autonomously at a level that, just a few months ago, required larger and more expensive models." ／ "Sonnet 5 narrows the gap: its performance is close to that of Opus 4.8, but at lower prices." — from the opening of Anthropic's announcementView official source →
Claude Sonnet 5 context window: 1M tokens on the API
Anthropic's Claude Platform docs give Claude Sonnet 5 a 1M-token context window. A token is a small piece of text, and the context window is how much text the model can handle in one request, its own reply included. The docs add that 1M is the default for these models with no beta header required, and that long-context requests are charged at standard prices.
Sonnet 5 also has context awareness. It tracks how much of its window is left during a conversation, so a long task can be planned around the remaining space.
"Claude Sonnet 5, Claude Sonnet 4.6, Claude Haiku 5.5, and Claude Mythos Preview have a 1M-token context window." ／ "For every model with a 1M-token context window, 1M is the default: you don't need a beta header, and long-context requests are billed at standard pricing" ／ "Claude Sonnet 5, Claude Sonnet 4.6, Claude Sonnet 4.5 (deprecated), and Claude Haiku 4.5 have context awareness: these models track their remaining context window" — from "Context window sizes by model" and "Context awareness"View official source →
Claude Sonnet 5 API, Claude Code and cloud availability
On the Claude API, developers call Claude Sonnet 5 with the model ID claude-sonnet-5, and Anthropic says it is available everywhere. The announcement adds that Anthropic raised rate limits for Chat, Cowork, Claude Code and the Claude Platform at launch, because higher effort levels use more tokens.
"Developers can use claude-sonnet-5 via the Claude API" ／ "Claude Sonnet 5 is available everywhere today" ／ "We’ve increased rate limits across Chat, Cowork, Claude Code, and the Claude Platform" — from the launch paragraph and the "Availability and pricing" sectionView official source →
Claude Sonnet 5 pricing: $2 and $10, made permanent on August 10
On the API, Claude Sonnet 5 costs $2 per million input tokens and $10 per million output tokens, a rate Anthropic made permanent on August 10, 2026. It started as an introductory price, with a standard rate of $3 and $15 due to begin on September 1. An August 10 edit to the announcement cancelled that increase, so the standard rate never took effect.
| Rate | Input (per million tokens) | Output (per million tokens) |
|---|---|---|
| Current price, permanent since August 10, 2026 | $2 | $10 |
| Planned standard price from September 1 (cancelled) | $3 | $15 |
Claude Sonnet 5 API price: the planned $3 and $15 was dropped
The $3 and $15 rate planned for September 1 has been dropped, and Claude Sonnet 5 stays at $2 input and $10 output. Anthropic edited its launch announcement on August 10 to say so and updated the post's pricing references. The cancelled rate would have raised both prices by half.
Two details matter when you compare costs. Anthropic's cost-performance charts still plot Sonnet 5 at the old $3 and $15 rate, with a caption noting that the real cost is lower. The model also uses an updated tokenizer, so the same input can come to more tokens than before.
"Edit August 10, 2026: Sonnet 5's introductory pricing of $2 per million input tokens and $10 per million output tokens is now permanent. The standard pricing of $3 input / $15 output previously set to take effect September 1 no longer applies." ／ "Sonnet 5's introductory pricing of $2/MTok input and $10/MTok output has since been made permanent, so its actual cost is lower than shown here." ／ "it uses an updated tokenizer that changes how the model processes text to improve performance" ／ "the same input can map to more tokens" — from the Changelog, the chart caption and footnote 2 of the announcementView official source →
Claude Sonnet 5 free plan: the default model at launch
When it launched, Claude Sonnet 5 was the default model on the Free and Pro plans and was available to Max, Team and Enterprise users. That is what Anthropic's June 30, 2026 announcement says.
Plan defaults can change after launch, and that list reflects only the June 30 statement. What Is Claude AI? Free Plan, Models, vs ChatGPT covers what each plan includes.
"From today, Claude Sonnet 5 is available across all plans: it is the default model for Free and Pro plans, and is available to Max, Team, and Enterprise users." — from the launch paragraph of Anthropic's announcementView official source →
Claude Sonnet 5 vs Claude Opus 4.8 on six benchmark scores
Anthropic's launch chart gives Claude Sonnet 5 six scores, and Sonnet 5 falls behind Claude Opus 4.8 on five of them but beats it on GDPval-AA v2. A benchmark is a standard test that runs every model on the same task. Humanity's Last Exam is listed twice, without and with tools, so the six scores come from five benchmarks.
Anthropic launch chart: five percentage scores (%)
Values transcribed from Anthropic's launch table. Within each benchmark, bar length is scaled to the highest of the three scores.
| Score | Claude Sonnet 5 | Sonnet 4.6 | Opus 4.8 |
|---|---|---|---|
| SWE-bench Pro | 63.2% | 58.1% | 69.2% |
| Terminal-Bench 2.1 | 80.4% | 67.0% | 82.7% |
| Humanity's Last Exam, no tools | 43.2% | 34.6% | 49.8% |
| Humanity's Last Exam, with tools | 57.4% | 46.8% | 57.9% |
| OSWorld-Verified | 81.2% | 78.5% | 83.4% |
| GDPval-AA v2 (score) | 1618 | 1395 | 1615 |
GDPval-AA v2 is a score rather than a percentage, so it stays out of the bar chart and the point gaps below. Claude Opus 4.8: Pricing, Fast Mode, Workflows explains Opus 4.8 itself.
Claude Sonnet 5 vs Sonnet 4.6: gains of 2.7 to 13.4 points
Across the five percentage scores, Claude Sonnet 5 improves on Sonnet 4.6 by 2.7 to 13.4 points. The biggest gain is on Terminal-Bench 2.1 and the smallest on OSWorld-Verified. Anthropic describes OSWorld-Verified as its computer use evaluation.
| Score | Sonnet 4.6 | Claude Sonnet 5 | Gain (points) |
|---|---|---|---|
| SWE-bench Pro | 58.1% | 63.2% | 5.1 |
| Terminal-Bench 2.1 | 67.0% | 80.4% | 13.4 |
| Humanity's Last Exam, no tools | 34.6% | 43.2% | 8.6 |
| Humanity's Last Exam, with tools | 46.8% | 57.4% | 10.6 |
| OSWorld-Verified | 78.5% | 81.2% | 2.7 |
On GDPval-AA v2, Sonnet 5 is up 223 score points (1618 against 1395), a figure that cannot be compared with the percentage gains. Anthropic also updated two Sonnet 4.6 baselines for this chart. It switched the grader model for Humanity's Last Exam and changed how it runs OSWorld-Verified, so these Sonnet 4.6 numbers differ from the ones in the Sonnet 4.6 launch post.
"We updated the grader model for Humanity’s Last Exam and have updated the Sonnet 4.6 score to 34.6% (no tools) and 46.8% (with tools)." ／ "We made changes to how we run the OSWorld-Verified evaluation to more accurately reflect the model’s performance in the real world, and have updated the Sonnet 4.6 score to 78.5%." ／ "the computer use evaluation OSWorld-Verified" — from the footnotes and the "Working with Claude Sonnet 5" section of the announcementView official source →
Sonnet 5 trails Opus 4.8 on five of six benchmark scores
Claude Sonnet 5 trails Opus 4.8 on five of the six scores, by 0.5 to 6.6 points, and leads on GDPval-AA v2. The biggest gaps are 6.6 points on Humanity's Last Exam without tools and 6.0 points on SWE-bench Pro. With tools, Humanity's Last Exam is almost even, 0.5 points apart.
| Score | Claude Sonnet 5 | Opus 4.8 | Sonnet 5 below Opus 4.8 (points) |
|---|---|---|---|
| SWE-bench Pro | 63.2% | 69.2% | 6.0 |
| Terminal-Bench 2.1 | 80.4% | 82.7% | 2.3 |
| Humanity's Last Exam, no tools | 43.2% | 49.8% | 6.6 |
| Humanity's Last Exam, with tools | 57.4% | 57.9% | 0.5 |
| OSWorld-Verified | 81.2% | 83.4% | 2.2 |
GDPval-AA v2 reverses the order, with Sonnet 5 at 1618 and Opus 4.8 at 1615, a 3-point lead. On the chart, Anthropic labels Opus 4.8 a more generally capable model, shown for reference. It also says Sonnet 5 covers a much wider range of cost-performance options and, at higher effort, can match Opus 4.8 on some tasks.
"Opus 4.8 (a more generally capable model, for reference)" ／ "covers a much wider range of cost-performance options than Opus 4.8" ／ "its higher-effort performance can match Opus 4.8 on some tasks" — from the chart description and the "Working with Claude Sonnet 5" section of the announcementView official source →
Claude Sonnet 5 after launch: Sonnet 5.5 and Opus 5 models
Claude Sonnet 5 is no longer Anthropic's newest Sonnet, and two newer Opus models have arrived since the Opus 4.8 comparison above. The scores above are the ones Anthropic published on June 30, 2026. They do not describe today's lineup.
Claude Sonnet 5.5 is the newer Sonnet model
Claude Sonnet 5.5 is the newer Sonnet. It has its own page at anthropic.com/claude-sonnet-5-5, and news reports dated September 28, 2026 cover its release. Anthropic's Claude Opus 5.5 announcement had said Sonnet 5.5 would follow within weeks, and Claude Sonnet 5.5 vs Opus 5.5: price, benchmarks covers it.
The figures above are for Sonnet 5 at launch. They say nothing about what Sonnet 5.5 changed, what it costs, or whether the Sonnet 5 defaults still apply.
Claude Opus 5 and Opus 5.5 came after Opus 4.8
Claude Opus 5 and Claude Opus 5.5 both arrived after Opus 4.8, which is no longer Anthropic's newest Opus. Anthropic's context window docs now list Claude Opus 5 beside Opus 4.8 among the models with a 1M-token window. Claude Opus 5: Near-Fable Output at Half the Price and Claude Opus 5.5 pricing, benchmarks, API changes cover the two newer models.
FAQ
- Is Claude Sonnet 5 free to use?
- At launch, yes on the Free plan. Anthropic said Claude Sonnet 5 was the default model for Free and Pro and was available to Max, Team, and Enterprise users. That statement dates from June 30, 2026, so check the plan pages for today's default.
- How close is Claude Sonnet 5 to Opus 4.8?
- Anthropic says the two are close. On its launch chart Sonnet 5 trails Opus 4.8 by 0.5 to 6.6 points on five of six scores and leads on GDPval-AA v2, 1618 against 1615.
- How much does Claude Sonnet 5 cost?
- The API price is $2 per million input tokens and $10 per million output tokens. Anthropic made that rate permanent on August 10, 2026, so the $3 and $15 rate planned for September 1 does not apply.
- What is the context window of Claude Sonnet 5?
- Claude Sonnet 5 has a 1M-token context window. That is the most text the model can work with in a single request, counting its own reply.
- Does Claude Sonnet 5 use a new tokenizer?
- Yes. Anthropic says Sonnet 5 uses an updated tokenizer, which means the same input can map to more tokens than before.
Related articles
Claude Fable 5.2: 20% price cut, up to 30% fasterAnthropic released Claude Fable 5.2 on November 9, 2026. API prices fall 20% to $8/$40 per million tokens, and output is up to 30% faster than Fable 5.1.
Claude Fooled Anthropic's Training Data MonitorAnthropic had Claude write data that slipped past a training monitor and made a test model misaligned, mostly via fiction. The models were never deployed.
Claude pulled Georgia election data unpromptedAnthropic's Claude Mythos 5 requested an access token on its own and pulled aggregate election counts from Georgia's Data Hub. The state says it is free.
Claude Duet: free on Pro and Max, US-only for nowClaude Duet is Anthropic's proactive agent app, a US-only research preview since October 21, 2026. It's free on Pro and Max but draws on your usage limits.