Skip to content

GPT-6.2 Astra · Writing

GPT-6.2 Astra for writing: our test runs

GPT-6.2 Astra 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.

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

Short answer

In our test runs on November 17, 2026, GPT-6.2 Astra passed 5 of 5 of our writing prompts (both hard ones): tied 1st of 27 models. Each reply counted as 1 message on Pro, where it has 31 a month, and cost $0.0087 to run on average.

Prompt by prompt

Each of the 5 writing prompts, and how GPT-6.2 Astra's reply did.

GPT-6.2 Astra on each writing prompt
PromptResultCostTime
Announce a second bakery shop on LinkedInEverydayPassed: Graded 5.0 of 5 on average (lowest 5).$0.00995.3 s
Rewrite corporate jargon in plain wordsEverydayPassed: Graded 4.3 of 5 on average (lowest 4).$0.00742.2 s
Decline a meeting and offer two timesEverydayPassed: Graded 5.0 of 5 on average (lowest 5).$0.00733.5 s
A product announcement with five rulesHardPassed: Graded 4.3 of 5 on average (lowest 4).$0.00754.9 s
Argue both sides of free busesHardPassed: Graded 4.7 of 5 on average (lowest 4).$0.01158.0 s

Every writing reply from GPT-6.2 Astra

  1. Announce a second bakery shop on LinkedIn

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

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

    On 14 November, we’re opening our second Crumb & Co. shop in Leeds. We’ll have the coffee on, and on opening day it’s free. If you’re nearby, we’d love you to pop in for a cup and say hello.
    
    One of the nicest parts of running a small bakery is getting to know people over the counter. To everyone who’s visited our first shop, told a friend about us or stopped for a chat on a busy morning: thank you. Your support has made this second shop possible.
    
    We’re looking forward to getting to know our new neighbours in Leeds. We hope to see you on 14 November!

    563 tokens in, 136 out · 5.3 s · $0.0099 · 1 message on Pro · answered by openai/gpt-6.2-astra via OpenAI ·

  2. Rewrite corporate jargon in plain words

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

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

    In the third quarter, teams from different departments worked together to put customers’ needs first. Among enterprise customers, retention measures improved by 12%, and the rate at which customers left fell by 0.3 percentage points. We also adjusted how we interacted with customers across all channels to make better use of our resources.

    591 tokens in, 68 out · 2.2 s · $0.0074 · 1 message on Pro · answered by openai/gpt-6.2-astra via OpenAI ·

  3. 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: Rescheduling the vendor demo
    
    Hi Tom,
    
    Thanks for the invitation. I can’t join the vendor demo on Wednesday at 11:00 because I’ll be presenting to the board. Would Thursday at 10:00 or Friday at 14:00 work for you instead?
    
    Best,  
    [Your name]

    569 tokens in, 69 out · 3.5 s · $0.0073 · 1 message on Pro · answered by openai/gpt-6.2-astra via OpenAI ·

  4. 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 desk lamp. Its arm turns 270 degrees, so you can direct light onto your keyboard or swing it over to a sketchbook. It dims in five steps and charges from USB-C.
    
    The Nordvik 2 costs €79 and ships on 1 December. What will you put in the spotlight?

    580 tokens in, 71 out · 4.9 s · $0.0075 · 1 message on Pro · answered by openai/gpt-6.2-astra via OpenAI ·

  5. Argue both sides of free buses

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

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

    A city should make its buses free because a fare that seems trivial to one rider can cost another a shift at work. A free bus lets people take a job, visit a clinic, or pick up a child without weighing the trip against the grocery bill. It also speeds boarding and ends disputes over payment. Everyone who depends on a bus arriving on time benefits when the driver can simply open the doors and let people on.
    
    A city should keep fares if replacing them would cost riders the service they need most. For a hospital cleaner whose shift ends after the last bus leaves, a free ride that never comes is no bargain. Use fare revenue to run buses later and more often, and offer free passes to people who cannot afford them. That gives the cleaner both an affordable ride and a way home, rather than paying for a discount for riders who need neither.

    558 tokens in, 175 out · 8.0 s · $0.0115 · 1 message on Pro · answered by openai/gpt-6.2-astra via OpenAI ·

How it compares on writing

GPT-6.2 Astra beside the job's picks among every model. 14 models share the hard-problems pick, GPT-6.2 Astra among them.

GPT-6.2 Astra beside other models on writing
ModelPassedOn ProCost per reply
Gemini 3.8 FlashOur pick: hard problems (shared), best value5 of 5250 a month on Pro$0.0005
DeepSeek V4.1 FlashOur pick: everyday4 of 560 a day on Pro$0.0003
GPT-6.2 AstraOur pick: hard problems (shared)5 of 531 a month on Pro$0.0087
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: Plus $3.75, 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.

GPT-6.2 Astra next to each model on writing

GPT-6.2 Astra 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 5.2: GPT-6.2 Astra answered 1.4× later (4.9 s to 3.5 s), cost 1.3× less, and passed 5 of 5 to its 5.

  • Against Claude Fable 5.1: GPT-6.2 Astra answered 1.2× later (4.9 s to 4.1 s), cost 1.5× less, and passed 5 of 5 to its 4. Only GPT-6.2 Astra passed Argue both sides of free buses.

  • Against Claude Opus 5.5: GPT-6.2 Astra answered within 10% of its time (4.9 s to 4.8 s), cost 1.8× less, and passed 5 of 5 to its 5.

  • Against Claude Sonnet 5.5: GPT-6.2 Astra answered 1.8× later (4.9 s to 2.7 s), cost 3.3× more, and passed 5 of 5 to its 4. Only GPT-6.2 Astra passed Argue both sides of free buses.

  • Against Claude Sonnet 5: GPT-6.2 Astra answered 1.4× later (4.9 s to 3.6 s), cost 2.3× more, and passed 5 of 5 to its 3. Only GPT-6.2 Astra passed Announce a second bakery shop on LinkedIn; A product announcement with five rules.

  • Against Claude Haiku 5.5: GPT-6.2 Astra answered 2.9× later (4.9 s to 1.7 s), cost 65.8× more, and passed 5 of 5 to its 3. Only GPT-6.2 Astra passed Rewrite corporate jargon in plain words; Argue both sides of free buses.

  • Against Claude Haiku 4.5: GPT-6.2 Astra answered 2.9× later (4.9 s to 1.7 s), cost 6.7× more, and passed 5 of 5 to its 4. Only GPT-6.2 Astra passed A product announcement with five rules.

  • Against GPT-6 Astra: GPT-6.2 Astra answered within 10% of its time (4.9 s to 4.7 s), cost 1.4× more, and passed 5 of 5 to its 4. Only GPT-6.2 Astra passed Rewrite corporate jargon in plain words.

  • Against GPT-6.2 Sol: GPT-6.2 Astra answered 2.7× later (4.9 s to 1.8 s), cost 3.9× more, and passed 5 of 5 to its 5.

  • Against GPT-6.1 Sol: GPT-6.2 Astra answered 1.1× sooner (4.9 s to 5.5 s), cost 4.6× more, and passed 5 of 5 to its 5.

  • Against GPT-6 Sol: GPT-6.2 Astra answered 1.5× later (4.9 s to 3.3 s), cost 6.6× more, and passed 5 of 5 to its 5.

  • Against GPT-6.1 Terra: GPT-6.2 Astra answered 1.3× later (4.9 s to 3.9 s), cost 13.3× more, and passed 5 of 5 to its 5.

  • Against GPT-6 Luna: GPT-6.2 Astra answered 1.9× later (4.9 s to 2.6 s), cost 155.9× more, and passed 5 of 5 to its 3. Only GPT-6.2 Astra passed Rewrite corporate jargon in plain words; Argue both sides of free buses.

  • Against Gemini 4 Carbon: GPT-6.2 Astra answered 4.3× sooner (4.9 s to 21.0 s), cost 1.4× more, and passed 5 of 5 to its 5.

  • Against Gemini 4 Argon: GPT-6.2 Astra answered 1.6× sooner (4.9 s to 7.9 s), cost 6.2× more, and passed 5 of 5 to its 4. Only GPT-6.2 Astra passed Argue both sides of free buses.

  • Against Gemini 4 Hydrogen: GPT-6.2 Astra answered 1.6× sooner (4.9 s to 7.9 s), cost 1.3× more, and passed 5 of 5 to its 5.

  • Against Gemini 3.1 Pro (preview): GPT-6.2 Astra answered 1.5× sooner (4.9 s to 7.4 s), cost 1.7× more, and passed 5 of 5 to its 3. Only GPT-6.2 Astra passed A product announcement with five rules; Argue both sides of free buses.

  • Against Gemini 3.8 Flash: GPT-6.2 Astra answered 1.6× later (4.9 s to 3.1 s), cost 17.8× more, and passed 5 of 5 to its 5.

  • Against DeepSeek V4.1 Flash: GPT-6.2 Astra answered 3.1× later (4.9 s to 1.6 s), cost 26.6× more, and passed 5 of 5 to its 4. Only GPT-6.2 Astra passed Rewrite corporate jargon in plain words.

  • Against DeepSeek V4 Pro: GPT-6.2 Astra answered 1.5× later (4.9 s to 3.3 s), cost 6.8× more, and passed 5 of 5 to its 5.

  • Against Grok 4.7: GPT-6.2 Astra answered 1.8× sooner (4.9 s to 8.6 s), cost 1.5× more, and passed 5 of 5 to its 5.

  • Against Kimi K3.1 Pro: GPT-6.2 Astra answered 1.4× sooner (4.9 s to 6.9 s), cost 1.8× more, and passed 5 of 5 to its 5.

  • Against Kimi K3: GPT-6.2 Astra answered within 10% of its time (4.9 s to 5.1 s), cost 2.7× more, and passed 5 of 5 to its 4. Only GPT-6.2 Astra passed Argue both sides of free buses.

  • Against GLM 5.3: GPT-6.2 Astra answered 3.2× later (4.9 s to 1.5 s), cost 10.2× more, and passed 5 of 5 to its 4. Only GPT-6.2 Astra passed Argue both sides of free buses.

  • Against GLM 5.3 Flash: GPT-6.2 Astra answered 4.8× later (4.9 s to 1.0 s), cost 53.4× more, and passed 5 of 5 to its 4. Only GPT-6.2 Astra passed A product announcement with five rules.

  • Against Mistral Large 4: GPT-6.2 Astra answered 1.5× sooner (4.9 s to 7.2 s), cost 5.9× more, and passed 5 of 5 to its 5.

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.

GPT-6.2 Astra, and writing, elsewhere

Questions

Is GPT-6.2 Astra good for writing?

In our test runs it passed 5 of 5 writing prompts, tied 1st of the 27 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 GPT-6.2 Astra use?

1 message each on Pro, where it has 31 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, GLM, and Mistral, in one chat.

See what a message costs before you send it. Every plan is month to month: cancel in two clicks. Sign in with an email link, no password.