What is Claude Sonnet 5?
Claude Sonnet 5 is Anthropic’s Sonnet-class model, released on June 30, 2026 and listed as Active (latest) on its official model page. Anthropic positions it as the best combination of speed and intelligence, and its announcement calls it the most agentic Sonnet model yet. The published specifications include a 1M-token context window, a 128K-token maximum output, text and image input, and a list price of $2 per million input tokens and $10 per million output tokens. Two details matter more than those headline numbers. Adaptive thinking is on by default, so a request that ran without thinking on Sonnet 4.6 will think on Sonnet 5, and the new tokenizer produces roughly 30% more tokens for the same text, which means a budget sized on an older Sonnet model no longer describes this one. Here you meet the model through an ordinary chat box: describe the task, attach an image, read a streaming answer and refine it in follow-up turns. This workspace does not expose every control Anthropic’s API offers, and the one control it does expose for this model — web search — is switched on here but not for every model on this site, so the sections below keep the official specifications and this site’s controls clearly apart.
What Claude Sonnet 5 is worth using for
Run a backlog of short tasks
Anthropic lists Sonnet 5 as Fast relative to the rest of the current lineup, and at $2 and $10 per million tokens it sits below the Opus tier on price. That combination suits routing work: send the many routine requests here and reserve a costlier model for the few tasks that actually need it. Effort is the lever inside a single model, and the low and medium settings are the ones to reach for when latency and token spend matter more than depth.
I will paste a list of small, independent tasks. Work through them one at a time and keep every answer short — a result, not an explanation. If any item needs information I have not given you, stop and ask instead of guessing.
Work through long documents and multi-step jobs
The context window is 1M tokens by default on the official API, and Anthropic bills the whole window at the standard per-token rate, so a very long request is not surcharged. Read that as an API specification rather than a promise about this page. Paste the passage you need when a file matters, because a PDF here is text-extracted first and a scan will fail. Because the new tokenizer produces about 30% more tokens for the same text, the same document also occupies more of any budget you estimated on an earlier Sonnet model.
I will paste a long document in sections. After each section, give me the decisions it contains, the action items with owners, and the open questions. Quote the sentence behind each decision, and mark anything the document leaves unspecified rather than filling it in yourself.
Read a dense screenshot or chart
Sonnet 5 takes text and images as input and returns text, so an interface screenshot or a dashboard is fair game. Say what you want read out of it, and ask for what is visible to be kept apart from what is inferred. Anthropic notes that this model follows instructions literally, so the more precisely you name the region you care about, the more reliable the reading tends to be.
Here is a screenshot of a dense dashboard. Describe its information hierarchy and navigation, and tell me which labels are too small for you to read reliably. Keep what you can see separate from what you are inferring.
Decide whether to migrate
Sonnet 5 is a drop-in upgrade from Sonnet 4.6 with three behaviour changes: adaptive thinking is on by default, manual extended thinking now returns a 400 error, and setting temperature, top_p or top_k to a non-default value also returns a 400 error. If you already run Sonnet 4.6, the practical question is not whether the newer model is stronger but what your token budgets and thinking settings were originally tuned for.
Help me judge whether moving to a newer Sonnet model is worth it. Ask me what I run today, how my token limits were sized, and whether I set thinking or sampling parameters. Then list what would break outright and what would simply cost more, and tell me what to test first.
Official Claude Sonnet 5 specifications
- Developer
- Anthropic
- Model ID
- claude-sonnet-5
- Official context window
- 1M tokens
- Official maximum output
- 128K tokens
- Official input and output
- Text and images in, text out
- Reliable knowledge cutoff
- January 2026
Claude Sonnet 5 and Claude Opus 5 side by side
Every value below comes from the same comparison table on Anthropic’s Claude Sonnet 5 model page, checked on September 20, 2026. They describe the two API models rather than this website, and the prices are Anthropic list prices in US dollars, not what this site charges in points.
| Specification | Claude Sonnet 5 | Claude Opus 5 |
|---|---|---|
| Context window | 1M tokens | 1M tokens |
| Max output | 128K tokens | 128K tokens |
| Price per MTok, in / out | $2 / $10 | $5 / $25 |
| Comparative latency | Fast | Moderate |
| Thinking | Adaptive | Adaptive |
| Default effort | high | high |
| Knowledge cutoff | Jan 2026 | May 2026 |
How to use Claude Sonnet 5 online
Choose the model
This page opens with Claude Sonnet 5 selected. The model navigation on the left and the selector in the chat box stay in sync, and switching models keeps the current conversation. Start a new chat when you want a clean context.
Describe the task and the format
State the goal, the constraints and the shape of the answer you want back. Sonnet 5 follows instructions literally, especially at lower effort, so an instruction aimed at one item will not quietly spread to the rest — say so explicitly when it should. Send one of the prompts from this guide as it is, or edit it first.
Sign in, send and check the cost
Sending a message needs an account with available points. The composer reads a reference price from the live model catalog, and a finished request reports what it actually cost. Follow up with corrections rather than starting over.
What Claude Sonnet 5 costs
Anthropic lists Claude Sonnet 5 at $2 per million input tokens and $10 per million output tokens, with cache writes at $2.50 per million for the five-minute window and $4 per million for the one-hour window. Cache reads cost $0.20 per million, which is 10% of the input price and the standard multiplier across Anthropic’s line. Batch requests are half price, and the full 1M-token context window is billed at the standard rate, so a 900k-token request costs the same per token as a 9k-token one. One price note carries dates. Anthropic launched Sonnet 5 at $2 and $10 as introductory pricing through August 31, 2026, and has since made those figures the standard price, so the increase to $3 and $15 that was scheduled for September 1, 2026 will not take effect. None of that is this site’s checkout price. Here you pay in points: the composer shows a reference price taken from the live catalog, and a finished request reports its actual cost. As of the information-check date below, that catalog entry lists its input and output reference prices in a 1 to 5 ratio and a cache-read row at about 10% of input — the same shape as Anthropic’s list prices, expressed in this site’s own unit rather than in dollars.
View points and plans ↗Claude Sonnet 5 FAQ
Can I use Claude Sonnet 5 for free?
Browsing this page and preparing a prompt costs nothing. Sending a message needs an account with available points, so Claude Sonnet 5 is not an unlimited free service here. Whether an account has trial points depends on the offer shown inside that account; unlimited free chat is not promised.
Do I need an Anthropic API key?
No. This page runs on your NanoBanana Pro account and point balance. If you want to call Anthropic directly, the official model page and pricing page linked below are the right starting points.
Can it read images and PDFs?
Anthropic lists text and images as the inputs for this model. This site accepts JPEG, PNG, WebP and PDF files up to 24 MB, but a PDF is not sent as a document: the text is extracted first, so a scanned page or a complex layout may come back as an error. Paste the passage manually in that case. Audio and video uploads are not available here.
Why is web search available but thinking disabled?
They are two switches with two different answers. Anthropic documents adaptive thinking that is on by default on this model and steered by an effort parameter, and the catalog entry this page runs on declares no reasoning level, so the composer disables that control. That same catalog entry does mark web search as supported, so the search toggle is enabled and you can use it. A greyed-out control in the chat box means unavailable, not broken.
How long can an answer be?
Anthropic documents a 1M-token context window and a 128K-token maximum output for the API model, plus up to 300K output tokens on the Message Batches API with a beta header. This site requests at most 4,096 output tokens per reply, which is a workspace setting rather than a model limit. Ask for long deliverables in parts, and read the context window as an API specification rather than a promise about a single conversation here.
What changes if I was already using Sonnet 4.6?
Anthropic calls Sonnet 5 a drop-in upgrade with three behaviour changes: adaptive thinking runs by default, manual extended thinking returns a 400 error, and setting temperature, top_p or top_k to a non-default value returns a 400 error. There is also a new tokenizer that produces roughly 30% more tokens for the same text, so token counts and max_tokens limits sized against Sonnet 4.6 are worth measuring again before you compare costs.
Can Claude Sonnet 5 generate images?
No. It takes text and images as input and returns text. You can ask it to describe an image or write a prompt for one, then generate the picture with an image model from the model library.
Is NanoBanana Pro the official Anthropic website?
No. NanoBanana Pro is an independent platform offering a shared workspace for several AI models. Anthropic develops Claude; the specifications and prices quoted on this page come from the official pages linked below.
Sources and model information
Information checked
Start with the task you run most often
Give it the goal, the constraints and the format you need, and keep the long jobs for the checkpoints that matter.
Chat with Claude Sonnet 5 ↑