Claude Sonnet 5.5 · Writing

Claude Sonnet 5.5 for writing: our test runs

Claude Sonnet 5.5 ran our 5 writing prompts in llmwise, through the same pipeline your messages take. Here's every reply as it came, whether it passed, what it cost and what it counts as on Pro, next to the other models.

Start free See pricing

Based on 5 of our test runs on September 28, 2026, through OpenRouter with the app's own prompt and settings. Updated September 28, 2026.

Short answer

In our test runs on September 28, 2026, Claude Sonnet 5.5 passed 4 of 5 of our writing prompts (1 of the 2 hard ones): tied 3rd of 16 models. Each reply counted as 1 message on Pro, where it has 125 a month, and cost $0.0034 to run on average.

Prompt by prompt

Each of the 5 writing prompts, and how Claude Sonnet 5.5's reply did.

Claude Sonnet 5.5 on each writing prompt
Prompt	Result	Cost	Time
Announce a second bakery shop on LinkedInEveryday	Passed: Graded 4.5 of 5 on average (lowest 4).	$0.0035	3.3 s
Rewrite corporate jargon in plain wordsEveryday	Passed: Graded 4.7 of 5 on average (lowest 4).	$0.0025	1.6 s
Decline a meeting and offer two timesEveryday	Passed: Graded 5.0 of 5 on average (lowest 5).	$0.0030	2.3 s
A product announcement with five rulesHard	Passed: Graded 4.3 of 5 on average (lowest 4).	$0.0030	2.3 s
Argue both sides of free busesHard	Failed: Graded 4.7 of 5 on average (lowest 4); but a paragraph of 98 words, over the 90 allowed.	$0.0051	4.4 s

Every writing reply from Claude Sonnet 5.5

Announce a second bakery shop on LinkedIn

An everyday prompt: “Announce a second bakery shop on LinkedIn”, in full.

Passed: Graded 4.5 of 5 on average (lowest 4).

We're opening a second shop, and we can hardly wait to tell you.

On 14 November, Crumb & Co. is coming to Leeds.

When we started out, it was a small bakery with a big oven and a simple idea: make good bread, treat people kindly, and let the rest follow. Your support over the years has made this next step possible, and we're grateful for every early morning, every loyal order and every kind word.

The new shop will have the same fresh loaves, pastries and cakes you know, baked with the same care. It will be a place to pop in, say hello and take something warm home.

To celebrate, we're giving away free coffee all day on opening day. Come along, bring a friend, and help us make our first day one to remember.

See you in Leeds on 14 November.

542 tokens in, 242 out · 3.3 s · $0.0035 · 1 message on Pro · answered by anthropic/claude-sonnet-5.5 via Anthropic · September 28, 2026

Rewrite corporate jargon in plain words

An everyday prompt: “Rewrite corporate jargon in plain words”, in full.

Passed: Graded 4.7 of 5 on average (lowest 4).

In Q3, teams from different departments worked together to put customers first. The result was a 12% increase in customer retention and a 0.30 percentage-point drop in the rate at which enterprise customers leave. We also reduced the number of ways we contact customers (such as email, phone, and chat) to free up staff time and resources.

601 tokens in, 102 out · 1.6 s · $0.0025 · 1 message on Pro · answered by anthropic/claude-sonnet-5.5 via Anthropic · September 28, 2026

Decline a meeting and offer two times

An everyday prompt: “Decline a meeting and offer two times”, in full.

Passed: Graded 5.0 of 5 on average (lowest 5).

**Subject:** Re: Vendor Demo on Wednesday

Hi Tom,

Thank you for inviting me to the vendor demo on Wednesday at 11:00. Unfortunately, I won't be able to attend, as I'll be presenting to the board at that time.

I'm still keen to see the demo, though. Would either of these work instead?

- Thursday at 10:00
- Friday at 14:00

If neither suits, let me know what else is available and I'll do my best to make it work.

Thanks again, and apologies for the inconvenience.

Best regards,
[Your name]

546 tokens in, 194 out · 2.3 s · $0.0030 · 1 message on Pro · answered by anthropic/claude-sonnet-5.5 via Anthropic · September 28, 2026

A product announcement with five rules

A hard prompt: “A product announcement with five rules”, in full.

Passed: Graded 4.3 of 5 on average (lowest 4).

Meet the Nordvik 2, the desk lamp designed for people who care how their workspace feels at every hour. Dim it in five steps, from bright focus light to a soft evening glow. Charge it from USB-C, so the same cable that powers your laptop keeps your lamp running. Its arm turns 270 degrees, letting you aim light exactly where you need it, whether you're sketching, reading, or typing late.

The Nordvik 2 costs €79 and ships on 1 December, just in time for the darker months.

Ready to give your desk the light it deserves?

558 tokens in, 185 out · 2.3 s · $0.0030 · 1 message on Pro · answered by anthropic/claude-sonnet-5.5 via Anthropic · September 28, 2026

Argue both sides of free buses

A hard prompt: “Argue both sides of free buses”, in full.

Failed: Graded 4.7 of 5 on average (lowest 4); but a paragraph of 98 words, over the 90 allowed.

Free buses make a city fairer, faster, and greener. Fares are a regressive tax on people who can least afford them, and dropping them instantly raises effective income for low-wage workers, students, and seniors. Boarding without payment cuts dwell time, so routes run quicker and more reliably. Cities that tried it, like Kansas City and Dunkirk, saw ridership jump, taking cars off congested streets and lowering emissions. Fare collection also eats money: machines, enforcement, and administration can consume a large share of what fares bring in. Free transit turns buses into a public good, like roads and libraries.

Free buses sound generous but quietly damage the service riders depend on. Fares provide a dependable revenue stream, and replacing it means higher taxes or budget cuts elsewhere, often leaving transit chronically underfunded. Money is better spent on frequency, reliability, and coverage, which matter more to low-income riders than a zero fare. Much of the new ridership comes from people who would have walked or cycled, not drivers, so climate gains are modest while crowding worsens. Targeted discounts help the poor directly without subsidizing every rider, including wealthy commuters, and keep agencies accountable to the people they serve.

522 tokens in, 401 out · 4.4 s · $0.0051 · 1 message on Pro · answered by anthropic/claude-sonnet-5.5 via Anthropic · September 28, 2026

How it compares on writing

Claude Sonnet 5.5 beside the job's picks among every model. 2 models share the hard-problems pick, and Claude Sonnet 5.5 isn't one of them.

Claude Sonnet 5.5 beside other models on writing
Model	Passed	On Pro	Cost per reply
GLM 5.3Our pick: best value	4 of 5	250 a month on Pro	$0.0006
GPT-6 LunaOur pick: hard problems (shared), everyday	5 of 5	60 a day on Pro	$0.0001
Claude Sonnet 5.5	4 of 5	125 a month on Pro	$0.0034

On Pro: Pro's count on each model. Cost: what OpenRouter charged us per reply, on average.

Every limit is published. Paid plans also have a monthly fair-use limit on AI cost: Pro $7.50, Max $20, Ultra $42, Studio $85. Using every message on your plan at typical sizes stays under it; very large messages and heavy research use it faster. Every limit, explained.

Claude Sonnet 5.5 next to each model on writing

Claude Sonnet 5.5 against each model on the same 5 writing prompts. One run each, through OpenRouter: a wait depends on the provider and the load that day, so a lead under 10% counts as close.

Against Claude Fable 5.1: Claude Sonnet 5.5 answered 2.5× sooner (2.3 s to 6.0 s), cost 5.1× less, and passed 4 of 5 to its 3. Only Claude Sonnet 5.5 passed A product announcement with five rules.
Against Claude Opus 5.5: Claude Sonnet 5.5 answered 3.5× sooner (2.3 s to 8.2 s), cost 5.4× less, and passed 4 of 5 to its 4.
Against Claude Sonnet 5: Claude Sonnet 5.5 answered 1.9× sooner (2.3 s to 4.5 s), cost within 10% of it, and passed 4 of 5 to its 4. Only Claude Sonnet 5.5 passed Announce a second bakery shop on LinkedIn. Only Claude Sonnet 5 passed Argue both sides of free buses.
Against Claude Haiku 4.5: Claude Sonnet 5.5 answered within 10% of its time (2.3 s to 2.2 s), cost 3.0× more, and passed 4 of 5 to its 4. Only Claude Sonnet 5.5 passed A product announcement with five rules. Only Claude Haiku 4.5 passed Argue both sides of free buses.
Against GPT-6 Astra: Claude Sonnet 5.5 answered 2.0× sooner (2.3 s to 4.8 s), cost 3.3× less, and passed 4 of 5 to its 5. Only GPT-6 Astra passed Argue both sides of free buses.
Against GPT-6 Sol: Claude Sonnet 5.5 answered 1.5× sooner (2.3 s to 3.5 s), cost 1.5× more, and passed 4 of 5 to its 4. Only Claude Sonnet 5.5 passed Rewrite corporate jargon in plain words. Only GPT-6 Sol passed Argue both sides of free buses.
Against GPT-6 Luna: Claude Sonnet 5.5 answered within 10% of its time (2.3 s to 2.4 s), cost 34.3× more, and passed 4 of 5 to its 5. Only GPT-6 Luna passed Argue both sides of free buses.
Against Gemini 3.1 Pro (preview): Claude Sonnet 5.5 answered 3.9× sooner (2.3 s to 9.1 s), cost 2.7× less, and passed 4 of 5 to its 4.
Against Gemini 3.8 Flash: Claude Sonnet 5.5 answered 1.7× sooner (2.3 s to 3.9 s), cost 2.6× more, and passed 4 of 5 to its 3. Only Claude Sonnet 5.5 passed Rewrite corporate jargon in plain words; A product announcement with five rules. Only Gemini 3.8 Flash passed Argue both sides of free buses.
Against DeepSeek V4.1 Flash: Claude Sonnet 5.5 answered 1.3× sooner (2.3 s to 3.1 s), cost 12.0× more, and passed 4 of 5 to its 4. Only Claude Sonnet 5.5 passed Rewrite corporate jargon in plain words. Only DeepSeek V4.1 Flash passed Argue both sides of free buses.
Against DeepSeek V4 Pro: Claude Sonnet 5.5 answered within 10% of its time (2.3 s to 2.2 s), cost 7.7× more, and passed 4 of 5 to its 3. Only Claude Sonnet 5.5 passed A product announcement with five rules.
Against Grok 4.7: Claude Sonnet 5.5 answered 1.7× sooner (2.3 s to 4.0 s), cost 1.2× less, and passed 4 of 5 to its 3. Only Claude Sonnet 5.5 passed Announce a second bakery shop on LinkedIn; Rewrite corporate jargon in plain words. Only Grok 4.7 passed Argue both sides of free buses.
Against Kimi K3: Claude Sonnet 5.5 answered 1.3× sooner (2.3 s to 3.0 s), cost 1.1× less, and passed 4 of 5 to its 4. Only Claude Sonnet 5.5 passed A product announcement with five rules. Only Kimi K3 passed Argue both sides of free buses.
Against GLM 5.3: Claude Sonnet 5.5 answered 1.3× later (2.3 s to 1.8 s), cost 5.9× more, and passed 4 of 5 to its 4.
Against GLM 5.3 Flash: Claude Sonnet 5.5 answered 3.6× sooner (2.3 s to 8.5 s), cost 10.6× more, and passed 4 of 5 to its 2. Only Claude Sonnet 5.5 passed Rewrite corporate jargon in plain words; A product announcement with five rules.

How these runs were done

Rubric (graded). The grader model scores the reply from 1 to 5 on each published criterion. It passes with an average of 4 or more and no criterion under 3, and only if it also meets the prompt's automatic rules (length, words it must or mustn't use).

The grader is Claude Opus 5.5 at low reasoning effort; its own replies are graded by GPT-6 Astra, so no model grades itself. Its prompt and every rubric are on the methods page.

How the runs were done, and every writing prompt.

Claude Sonnet 5.5, and writing, elsewhere

Questions

Is Claude Sonnet 5.5 good for writing?

In our test runs it passed 4 of 5 writing prompts, tied 3rd of the 16 models in llmwise. Every reply is on this page, so you can judge them yourself.

How many of my messages does a writing reply from Claude Sonnet 5.5 use?

1 message each on Pro, where it has 125 a month on Pro. The price of a message is fixed and shown before you send it, however long the reply.

How were these runs done?

The same way for every model: each prompt sent through llmwise's own pipeline, each reply checked the same way. The methods page has every prompt and how each is scored.

Claude, GPT, Gemini, DeepSeek, Grok, Kimi, and GLM, in one chat.

See what a message costs before you send it. Free is 5 messages to try; sign in with an email link, no password or card.

Start free See pricing