ChatGPT vs DeepSeek for writing

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.

Writing: GPT-6 Luna passed 5 of 5 and DeepSeek V4 Pro 5 of 5. GPT-6 Luna answered 1.6× sooner at the median, 2.4 s against 3.9 s. GPT-6 Luna cost 10.8× less, $0.0005 against $0.0054 for the 5 replies. Their replies ran to about the same length.

1. 1. Announce a second bakery shop on LinkedIn

   Everyday. GPT: 4 of 4 models passed, GPT-6 Luna the cheapest at $0.0001 · DeepSeek: 2 of 2 models passed, DeepSeek V4.1 Flash the cheapest at $0.0004. The prompt in full.

   GPT-6 Luna and DeepSeek V4 Pro both passed this one. GPT-6 Luna answered sooner (2.9 s to 3.9 s), and GPT-6 Luna's reply cost less ($0.0001 to $0.0010); GPT-6 Luna wrote 149 tokens of reply, DeepSeek V4 Pro 164.

   • GPT-6 Luna

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

     We’re delighted to share that Crumb & Co. is opening a second shop in Leeds on 14 November!
     
     What began as one small bakery has grown thanks to the lovely people who’ve stopped by, shared a loaf, and made us feel part of the community. We’re excited to bring our freshly baked bread, pastries, and warm welcomes to a new neighbourhood.
     
     To celebrate opening day, we’ll be giving away free coffee. Pop in, say hello, and enjoy a cup with us—we’d love to meet you.
     
     Thank you to everyone who has supported Crumb & Co. so far. We can’t wait to open the doors and start this next chapter together. See you in Leeds on 14 November!

     389 tokens in, 149 out · 2.9 s · $0.0001 · 1 message on Pro · answered by openai/gpt-6-luna via OpenAI · September 27, 2026

   • DeepSeek V4 Pro

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

     Big news from Crumb & Co. — we're opening a second shop, and this time we're heading to Leeds.
     
     Our new bakery will open its doors on Thursday 14 November, and we'd love you to be among the first through them. Come and see the new space, smell the bread coming out of the oven, and say hello to the team.
     
     To mark the day, we're giving away free coffee to everyone who visits — our way of saying thank you for supporting a small, independent bakery.
     
     We've spent months getting this little shop ready, and we're equal parts nervous and excited. If you're in Leeds, we'd be glad to see you. Bring a friend, grab a warm loaf, and help us make it feel like home.
     
     See you on the 14th.

     400 tokens in, 199 out (35 of them reasoning) · 3.9 s · $0.0010 · 1 message on Pro · answered by deepseek/deepseek-v4-pro-0813 via Wafer · October 7, 2026

2. 2. Rewrite corporate jargon in plain words

   Everyday. GPT: 2 of 4 models passed, GPT-6 Luna the cheapest at $0.0001 · DeepSeek: 1 of 2 models passed, DeepSeek V4 Pro at $0.0024. The prompt in full.

   GPT-6 Luna and DeepSeek V4 Pro both passed this one. GPT-6 Luna answered sooner (2.4 s to 5.8 s), and GPT-6 Luna's reply cost less ($0.0001 to $0.0024); GPT-6 Luna wrote 65 tokens of reply, DeepSeek V4 Pro 58.

   • GPT-6 Luna

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

     In Q3, teams across the company worked together to focus more on customers. Retention measures improved by 12%, and the rate of customer loss fell by 0.3 percentage points in the enterprise segment. The company also adjusted its customer contact channels to make better use of available capacity.

     417 tokens in, 124 out (59 of them reasoning) · 2.4 s · $0.0001 · 1 message on Pro · answered by openai/gpt-6-luna via OpenAI · September 27, 2026

   • DeepSeek V4 Pro

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

     In the third quarter, our teams worked together to focus more on customers. This produced a 12% increase in customer retention and a 0.30 percentage point drop in churn among business customers. We also fine-tuned our various customer contact channels to use resources more efficiently.

     431 tokens in, 576 out (518 of them reasoning) · 5.8 s · $0.0024 · 1 message on Pro · answered by deepseek/deepseek-v4-pro-0813 via Wafer · October 7, 2026

3. 3. Decline a meeting and offer two times

   Everyday. GPT: 4 of 4 models passed, GPT-6 Luna the cheapest at $0.0001 · DeepSeek: 2 of 2 models passed, DeepSeek V4.1 Flash the cheapest at $0.0002. The prompt in full.

   GPT-6 Luna and DeepSeek V4 Pro both passed this one. GPT-6 Luna answered sooner (1.3 s to 3.3 s), and GPT-6 Luna's reply cost less ($0.0001 to $0.0006); GPT-6 Luna wrote 64 tokens of reply, DeepSeek V4 Pro 82.

   • GPT-6 Luna

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

     Subject: Vendor Demo
     
     Hi Tom,
     
     Thanks for the invitation. I’m presenting to the board on Wednesday at 11:00, so I can’t make the demo then. Would Thursday at 10:00 or Friday at 14:00 work instead?
     
     Best,  
     [Your Name]

     395 tokens in, 64 out · 1.3 s · $0.0001 · 1 message on Pro · answered by openai/gpt-6-luna via OpenAI · September 27, 2026

   • DeepSeek V4 Pro

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

     Subject: Vendor demo – alternative times
     
     Hi Tom,
     
     Thanks for the invitation to the vendor demo on Wednesday at 11:00. Unfortunately, I’m presenting to the board at that time, so I won’t be able to join.
     
     Would Thursday at 10:00 or Friday at 14:00 work for you instead? I’m happy to make either time work.
     
     Best regards

     406 tokens in, 120 out (38 of them reasoning) · 3.3 s · $0.0006 · 1 message on Pro · answered by deepseek/deepseek-v4-pro-0813 via Wafer · October 8, 2026

4. 4. A product announcement with five rules

   Hard. GPT: 4 of 4 models passed, GPT-6 Luna the cheapest at $0.0001 · DeepSeek: 2 of 2 models passed, DeepSeek V4.1 Flash the cheapest at $0.0003. The prompt in full.

   GPT-6 Luna and DeepSeek V4 Pro both passed this one. GPT-6 Luna answered sooner (1.5 s to 1.8 s), and GPT-6 Luna's reply cost less ($0.0001 to $0.0005); GPT-6 Luna wrote 73 tokens of reply, DeepSeek V4 Pro 71.

   • GPT-6 Luna

     Passed: Graded 4.0 of 5 on average (lowest 3).

     Meet the Nordvik 2, a desk lamp designed to bring flexible light to your workspace. Choose from five brightness levels, charge it via USB-C, and turn its arm up to 270 degrees to direct light where you need it. The Nordvik 2 costs €79 and ships on 1 December. Ready to brighten your desk?

     406 tokens in, 73 out · 1.5 s · $0.0001 · 1 message on Pro · answered by openai/gpt-6-luna via OpenAI · September 27, 2026

   • DeepSeek V4 Pro

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

     Introducing the Nordvik 2 desk lamp. It dims in five steps, so you can find the right light for any task. It charges over USB-C, keeping your desk free of extra cables. The arm turns 270 degrees for precise positioning. Available for €79, with shipping on 1 December. Have you reserved yours yet?

     418 tokens in, 96 out (25 of them reasoning) · 1.8 s · $0.0005 · 1 message on Pro · answered by deepseek/deepseek-v4-pro-0813 via Wafer · October 8, 2026

5. 5. Argue both sides of free buses

   Hard. GPT: 4 of 4 models passed, GPT-6 Luna the cheapest at $0.0001 · DeepSeek: 2 of 2 models passed, DeepSeek V4 Pro the cheapest at $0.0009. The prompt in full.

   GPT-6 Luna and DeepSeek V4 Pro both passed this one. GPT-6 Luna answered sooner (2.5 s to 4.1 s), and GPT-6 Luna's reply cost less ($0.0001 to $0.0009); GPT-6 Luna wrote 186 tokens of reply, DeepSeek V4 Pro 153.

   • GPT-6 Luna

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

     Making buses free would give every resident reliable access to jobs, school, health care, and daily necessities, regardless of income. It would also speed boarding, reduce fare-enforcement costs, and encourage people to leave cars at home—easing congestion, pollution, and parking pressure. A city that treats public transit as essential infrastructure, rather than a service people must qualify to afford, can build a more connected and equitable community.
     
     But “free” buses still have to be paid for, and the bill could divert money from more urgent improvements: frequent service, safer stops, cleaner vehicles, and routes to underserved neighborhoods. If buses remain slow, crowded, or unreliable, removing fares may do little to persuade drivers to switch, while reducing funds that could make transit genuinely useful. Targeted fare assistance can protect riders who need it most without subsidizing every trip, including those by people who can readily afford to pay.

     384 tokens in, 186 out · 2.5 s · $0.0001 · 1 message on Pro · answered by openai/gpt-6-luna via OpenAI · September 27, 2026

   • DeepSeek V4 Pro

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

     Making buses free turns public transit into a true public good, like sidewalks or libraries. It removes fares as a barrier for low-income riders, boosts ridership, cuts traffic congestion and emissions, and speeds boarding by eliminating payment delays. Cities already subsidize most transit costs; dropping the small fare share is a modest investment with outsized returns in mobility, equity, and cleaner air.
     
     The strongest case against free buses is that fares, however small, are the only direct signal of demand. Removing them invites overcrowding on popular routes while doing nothing to expand service where it is sparse. Scarce transit dollars are better spent on frequency, reliability, and coverage—improvements that actually attract riders—than on subsidizing trips for people who can already afford to pay.

     395 tokens in, 193 out (40 of them reasoning) · 4.1 s · $0.0009 · 1 message on Pro · answered by deepseek/deepseek-v4-pro-0813 via Wafer · October 8, 2026

• GPT-6 Astra vs DeepSeek V4 Pro: 5 of 5 each; DeepSeek V4 Pro answered 1.2× sooner at the median and cost 10.4× less.

• GPT-6 Astra vs DeepSeek V4.1 Flash: GPT-6 Astra 5 of 5, DeepSeek V4.1 Flash 4; DeepSeek V4.1 Flash answered 3.6× sooner at the median and cost 24.7× less.

• GPT-6.1 Sol vs DeepSeek V4 Pro: GPT-6.1 Sol 4 of 5, DeepSeek V4 Pro 5; GPT-6.1 Sol answered 1.2× sooner at the median and cost about the same.

• GPT-6.1 Sol vs DeepSeek V4.1 Flash: 4 of 5 each; DeepSeek V4.1 Flash answered 2.4× sooner at the median and cost 2.5× less.

• GPT-6 Sol vs DeepSeek V4 Pro: GPT-6 Sol 4 of 5, DeepSeek V4 Pro 5; GPT-6 Sol answered 1.1× sooner at the median and DeepSeek V4 Pro cost 2.1× less. GPT-6 Sol vs DeepSeek V4 Pro, on every job.

• GPT-6 Sol vs DeepSeek V4.1 Flash: 4 of 5 each; DeepSeek V4.1 Flash answered 2.6× sooner at the median and cost 5.0× less.

• GPT-6 Luna vs DeepSeek V4 Pro: 5 of 5 each; GPT-6 Luna answered 1.6× sooner at the median and cost 10.8× less. GPT-6 Luna vs DeepSeek V4 Pro, on every job.

• GPT-6 Luna vs DeepSeek V4.1 Flash: GPT-6 Luna 5 of 5, DeepSeek V4.1 Flash 4; DeepSeek V4.1 Flash answered 1.8× sooner at the median and GPT-6 Luna cost 4.5× less. GPT-6 Luna vs DeepSeek V4.1 Flash, on every job.

Every GPT and DeepSeek reply above was checked the same way as every other model's, by the rules published with the prompts: how the writing prompts are scored, and each one in full.

• The lineups

  GPT: 4 models, GPT-6 Astra, GPT-6.1 Sol, GPT-6 Sol, and GPT-6 Luna. DeepSeek: 2 models, DeepSeek V4 Pro and DeepSeek V4.1 Flash.

• Price per message

  The least expensive GPT model is GPT-6 Luna (60 messages a day on Pro); the least expensive DeepSeek model is DeepSeek V4.1 Flash (60 messages a day on Pro).

• Context window

  GPT goes up to 1.05M tokens (GPT-6 Astra); DeepSeek up to 1.05M tokens (DeepSeek V4 Pro).

• Images and PDFs

  DeepSeek V4 Pro doesn't read images. DeepSeek V4 Pro and DeepSeek V4.1 Flash get a PDF's text rather than the file itself.

• On the Free plan

  None of these models is on the Free plan. Paid plans have them, with messages every month: GPT-6 Astra from Pro up.

In llmwise, a message to GPT goes to its maker, OpenAI, or through OpenRouter when llmwise can't reach the maker directly. DeepSeek models are served only through OpenRouter, by endpoints that don't store or train on prompts. The Privacy Policy has the details.