Updated

Anthropic
Claude Sonnet 5
Anthropic's most agentic Sonnet at launch, close to Opus 4.8 and able to reply with thinking off.
What is Claude Sonnet 5?
Claude Sonnet 5 is a Sonnet model from Anthropic, released on June 30, 2026.
Anthropic called it the most agentic Sonnet yet: it plans, uses tools such as browsers and terminals and works on its own, with performance close to Claude Opus 4.8 at a lower price. It improved on Sonnet 4.6 in reasoning, tool use, coding and knowledge work, and its updated tokenizer can count the same input as up to 1.35 times as many tokens. Anthropic now lists it as a legacy model and recommends Claude Sonnet 5.5, but Sonnet 5 can still run with thinking off.
- Provider
- Anthropic
- Context window
- 1M tokens
- Input
- Text, Image
- Reasoning
- Off to max
- Max output
- 128K tokens
- Knowledge cutoff
- January 2026
- Released
- Jun 30, 2026
How Claude Sonnet 5 performs
Published by Anthropic
Results Anthropic reported for Claude Sonnet 5. Each lab runs its own setup, so compare these with other models only inside the same table.
| Benchmark | Score | Setting |
|---|---|---|
| SWE-bench Pro | 63.2% | |
| Terminal-Bench 2.1 | 80.4% | |
| Humanity's Last Exam | 43.2% | no tools |
| Humanity's Last Exam | 57.4% | with tools |
| OSWorld-Verified | 81.2% |
Source: anthropic.com
On LMArena
People compare two anonymous answers and vote for the better one. Ranked here against the other chat models you can pick in Super Agent.
Text
#18 of 27 in Super Agent
All text prompts, with style control on. #58 of 418 on the full leaderboard, from 51,784 votes.
- 1Claude Opus 5.51509
- 2Claude Fable 51503
- 3Claude Fable 5.11503
- 4Claude Opus 51491
- 5Gemini 3.7 Flash1490
- 18Claude Sonnet 51462
Coding
#14 of 27 in Super Agent
Text prompts that ask for code. #37 of 413 on the full leaderboard, from 13,777 votes.
- 1Claude Fable 51551
- 2Claude Opus 5.51548
- 3GPT-6 Astra1543
- 4Kimi K31539
- 5Claude Fable 5.11538
- 14Claude Sonnet 51520
WebDev
#19 of 27 in Super Agent
Building working web apps in Code Arena. #47 of 149 on the full leaderboard, from 16,126 votes.
- 1Claude Opus 5.51807
- 2GPT-6 Astra1775
- 3Claude Sonnet 5.51768
- 4GPT-6.1 Sol1749
- 5Claude Fable 5.11743
- 19Claude Sonnet 51541
Math
#19 of 27 in Super Agent
Text prompts with math problems. #50 of 404 on the full leaderboard, from 2,385 votes.
- 1Claude Fable 5.11529
- 2Claude Fable 51521
- 3Claude Opus 51517
- 4GPT-6.1 Sol1513
- 5Claude Opus 5.51512
- 19Claude Sonnet 51473
Hard prompts
#17 of 27 in Super Agent
Complex, multi-part and demanding text prompts. #49 of 418 on the full leaderboard, from 33,975 votes.
- 1Claude Opus 5.51535
- 2Claude Fable 51529
- 3Claude Fable 5.11523
- 4Kimi K31516
- 5Claude Opus 51515
- 17Claude Sonnet 51490
Creative writing
#20 of 27 in Super Agent
Stories, poems and other original writing. #64 of 416 on the full leaderboard, from 10,786 votes.
- 1Claude Opus 5.51520
- 2Claude Fable 51501
- 3Gemini 3.7 Flash1492
- 4Claude Fable 5.11487
- 5Gemini 3.1 Pro Preview1481
- 20Claude Sonnet 51435
Vision
#20 of 23 in Super Agent
Prompts that include images, with style control on. #44 of 167 on the full leaderboard, from 14,034 votes.
- 1Claude Fable 51309
- 2Gemini 3.7 Flash1295
- 3GPT-6.1 Sol1294
- 4Claude Opus 5.51293
- 5Claude Opus 51288
- 20Claude Sonnet 51262
Source: LMArena, leaderboard published . Scores move as new votes come in.
What you can do with Claude Sonnet 5
Thinking can be switched off for instant replies
Close to Claude Opus 4.8 at a lower cost, according to Anthropic
Plans and uses tools such as browsers and terminals
1M-token context window with image input
How to prompt Claude Sonnet 5
- Switch thinking off for quick rewrites and ideas; turn it on for planning.
- Attach reference images so it can describe them in the prompts it writes.
- If you want Anthropic's newest Sonnet, compare the same prompt on Sonnet 5.5.
Where Claude Sonnet 5 fits best
Dependable everyday agent work
Fast replies with thinking off
Scripts, captions and prompt writing
Reviewing images you attach
How Claude Sonnet 5 compares
| Model | LMArena text | LMArena coding | Context window | Reasoning |
|---|---|---|---|---|
| Claude Sonnet 5 | 1462#58 | 1520#37 | 1M tokens | Off to max |
| Kimi K2.6 | 1461#59 | 1516#44 | 256K tokens | On or off |
| GPT-6 Sol | 1459#62 | 1519#38 | 1.05M tokens | Off to max |
| DeepSeek V4 Pro | 1458#64 | 1501#79 | 1M tokens | Off, medium, high, extra high |
| Grok 4.5 | 1466#52 | 1514#48 | 500K tokens | Low, medium, high |
Good to know before you start
- Anthropic lists it as legacy and points to Claude Sonnet 5.5
- Its tokenizer can turn the same text into up to 1.35 times the tokens Sonnet 4.6 used
- Writes the prompts but renders nothing itself; Flashloop's generation models make the images and clips
Start chatting with Claude Sonnet 5
Chat with Claude Sonnet 5 in Super AgentSelect Claude Sonnet 5
Open Super Agent on Flashloop and choose Claude Sonnet 5 in the model picker.
Enter your prompt
Ask for what you need: an answer, a plan, or the images and videos you want the agent to make.
Send and keep going
Send the message. Claude Sonnet 5 answers in Super Agent and runs any generations you asked for.
Questions about Claude Sonnet 5
Can Claude Sonnet 5 reply without thinking?
Yes. In Super Agent you can set reasoning to off, or pick low, medium, high, extra high or max. It starts at low.
How does Claude Sonnet 5 compare with Sonnet 5.5?
Anthropic calls Sonnet 5.5 a clear upgrade that runs more than 30% faster. In Super Agent, Sonnet 5 keeps a thinking-off setting that Sonnet 5.5 doesn't offer.
What context window does Claude Sonnet 5 have?
1M tokens. It accepts text and images.
Is Claude Sonnet 5 still supported?
Anthropic lists it as a legacy model and recommends moving to Sonnet 5.5. You can still pick it in Super Agent's model picker.
More chat models on Flashloop
Kimi K2.6Moonshot AI's open coding model that turns prompts and visual inputs into working interfaces.
GPT-6 SolOpenAI's GPT-6 model between Luna and Astra, trained like Astra and able to reply with reasoning off.
DeepSeek V4 ProDeepSeek's largest V4 model for coding, knowledge and STEM; Super Agent runs the April preview.
Grok 4.5xAI's Grok 4.5, trained alongside Cursor for coding and office work, served at 80 tokens a second.