Skip to content

Comparison · Math

ChatGPT vs DeepSeek for math

On llmwise Pro, GPT-6.1 Sol gets up to 125 messages a month and DeepSeek V4 Pro up to 250 messages a month. ChatGPT is OpenAI's own app for its GPT models; llmwise has GPT models in its own chat, not ChatGPT itself. We ran the same 5 math prompts on all 5 GPT models and all 2 DeepSeek models and published every reply: the results, then GPT-6.1 Terra against DeepSeek V4.1 Flash prompt by prompt, then how GPT and DeepSeek compare on price per message, context and files.

Based on 35 of our test runs on , through OpenRouter with the app's own prompt and settings. Updated .

Short answer

In our math test runs on October 27, 2026, GPT's 5 models passed 24 of 25; GPT-6 Astra, GPT-6.1 Sol and 2 more each passed 5 of 5. DeepSeek's 2 models passed 10 of 10; DeepSeek V4 Pro and DeepSeek V4.1 Flash each passed 5 of 5. GPT-6.1 Terra and DeepSeek V4.1 Flash each passed 5 of the 5 prompts, so these math prompts don't split GPT and DeepSeek; the replies on this page show how they differ.

GPT and DeepSeek on our math test runs

Every GPT and DeepSeek model in llmwise on our 5 math prompts: how many replies passed, what each counted as on Pro, and what it cost to run.

GPT and DeepSeek on our math test runs
ModelPassedHard onesMessages used on ProCost per replyTime per reply
GPT-6 AstraOpenAI5 of 52 of 21 each, of 31 a month on Pro$0.00852.7 s
GPT-6.1 SolOpenAI5 of 52 of 21 each, of 125 a month on Pro$0.00092.0 s
GPT-6 SolOpenAI5 of 52 of 21 each, of 125 a month on Pro$0.00182.2 s
GPT-6.1 TerraOpenAI5 of 52 of 21 each, of 250 a month on Pro$0.00052.1 s
GPT-6 LunaOpenAI4 of 52 of 21 each, of 60 a day on Pro$0.00012.4 s
DeepSeek V4 ProDeepSeek5 of 52 of 21 each, of 250 a month on Pro$0.00092.9 s
DeepSeek V4.1 FlashDeepSeek5 of 52 of 21 each, of 60 a day on Pro$0.00021.1 s
GPT: 24 of 25 replies passed; DeepSeek: 10 of 10 replies passed. Messages used is what each reply counted as on Pro; cost is what OpenRouter charged us per reply.

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 vs DeepSeek, prompt by prompt

Each math prompt with GPT-6.1 Terra and DeepSeek V4.1 Flash side by side, as they replied: for each side, the model that passed the most math prompts (then the most hard ones, then the cheaper).

Math: GPT-6.1 Terra passed 5 of 5 and DeepSeek V4.1 Flash 5 of 5. DeepSeek V4.1 Flash answered 2.0× sooner at the median, 1.9 s against 0.9 s. DeepSeek V4.1 Flash cost 2.1× less, $0.0027 against $0.0012 for the 5 replies. DeepSeek V4.1 Flash's replies ran 20% longer, in tokens of reply, thinking not counted.

  1. 1. A discount, then sales tax

    Everyday. GPT: 5 of 5 models passed, GPT-6 Luna the cheapest at $0.0001 · DeepSeek: 2 of 2 models passed, DeepSeek V4.1 Flash the cheapest at $0.0001. The prompt in full.

    GPT-6.1 Terra and DeepSeek V4.1 Flash both passed this one. DeepSeek V4.1 Flash answered sooner (0.7 s to 1.2 s), and DeepSeek V4.1 Flash's reply cost less ($0.0001 to $0.0004); GPT-6.1 Terra wrote 38 tokens of reply, DeepSeek V4.1 Flash 41.

    • GPT-6.1 Terra

      Passed: Final answer 66: right.

      Discounted price: $80 × 0.75 = $60  
      With tax: $60 × 1.10 = $66
      
      Final answer: $66

      574 tokens in, 38 out · 1.2 s · $0.0004 · 1 message on Pro · answered by openai/gpt-6.1-terra via OpenAI ·

    • DeepSeek V4.1 Flash

      Passed: Final answer 66: right.

      $80 × 0.75 = $60 (after 25% discount)
      $60 × 1.10 = $66 (after 10% tax)
      
      Final answer: $66

      440 tokens in, 63 out (22 of them reasoning) · 0.7 s · $0.0001 · 1 message on Pro · answered by deepseek/deepseek-v4.1-flash via Together ·

  2. 2. Pens at 3 for $4

    Everyday. GPT: 4 of 5 models passed, GPT-6.1 Terra the cheapest at $0.0006 · DeepSeek: 2 of 2 models passed, DeepSeek V4.1 Flash the cheapest at $0.0004. The prompt in full.

    GPT-6.1 Terra and DeepSeek V4.1 Flash both passed this one. DeepSeek V4.1 Flash answered sooner (0.9 s to 3.1 s), and DeepSeek V4.1 Flash's reply cost less ($0.0004 to $0.0006); GPT-6.1 Terra wrote 84 tokens of reply, DeepSeek V4.1 Flash 146.

    • GPT-6.1 Terra

      Passed: Final answer 3 deals and 1 single pen, for 13.50.: right.

      A 3-pen deal costs $4, less than the $4.50 cost of three single pens. So buy as many 3-pen deals as possible:
      
      3 deals (9 pens) + 1 single pen = $12 + $1.50 = $13.50.
      
      Final answer: 3 deals and 1 single pen, for $13.50.

      576 tokens in, 113 out (29 of them reasoning) · 3.1 s · $0.0006 · 1 message on Pro · answered by openai/gpt-6.1-terra via OpenAI ·

    • DeepSeek V4.1 Flash

      Passed: Final answer Buy 3 packs of 3 (9 pens) plus 1 single pen, for 13.50.: right.

      3-pack rate: $4/3 ≈ $1.33 per pen, cheaper than $1.50 singles.
      
      Options:
      - 3 packs (9 pens) + 1 single = $12 + $1.50 = $13.50
      - 2 packs (6) + 4 singles = $8 + $6 = $14.00
      - 1 pack + 7 singles = $4 + $10.50 = $14.50
      - 10 singles = $15.00
      
      Best is 3 packs plus 1 single pen.
      
      Final answer: Buy 3 packs of 3 (9 pens) plus 1 single pen, for $13.50.

      442 tokens in, 291 out (145 of them reasoning) · 0.9 s · $0.0004 · 1 message on Pro · answered by deepseek/deepseek-v4.1-flash via Together ·

  3. 3. Compound interest over three years

    Everyday. GPT: 5 of 5 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-6.1 Terra and DeepSeek V4.1 Flash both passed this one. DeepSeek V4.1 Flash answered sooner (2.1 s to 2.6 s), and DeepSeek V4.1 Flash's reply cost less ($0.0002 to $0.0005); GPT-6.1 Terra wrote 57 tokens of reply, DeepSeek V4.1 Flash 72.

    • GPT-6.1 Terra

      Passed: Final answer 315.25: right.

      Balance after 3 years: $2,000 × 1.05³ = $2,315.25  
      Interest earned: $2,315.25 − $2,000 = $315.25
      
      Final answer: $315.25

      576 tokens in, 85 out (28 of them reasoning) · 2.6 s · $0.0005 · 1 message on Pro · answered by openai/gpt-6.1-terra via OpenAI ·

    • DeepSeek V4.1 Flash

      Passed: Final answer 315.25: right.

      A = P(1 + r)^n = 2000(1.05)^3
      
      1.05³ = 1.157625
      2000 × 1.157625 = 2315.25
      Interest = 2315.25 − 2000 = 315.25
      
      Final answer: $315.25

      441 tokens in, 111 out (39 of them reasoning) · 2.1 s · $0.0002 · 1 message on Pro · answered by deepseek/deepseek-v4.1-flash via Together ·

  4. 4. Four-digit numbers whose digits sum to 9

    Hard. GPT: 5 of 5 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-6.1 Terra and DeepSeek V4.1 Flash both passed this one. DeepSeek V4.1 Flash answered sooner (0.7 s to 1.7 s), and DeepSeek V4.1 Flash's reply cost less ($0.0003 to $0.0006); GPT-6.1 Terra wrote 103 tokens of reply, DeepSeek V4.1 Flash 105.

    • GPT-6.1 Terra

      Passed: Final answer 165: right.

      Let the digits be \(a,b,c,d\), where \(a\ge 1\). Set \(a'=a-1\). Then
      
      \[
      a'+b+c+d=8.
      \]
      
      By stars and bars, the number of nonnegative integer solutions is
      
      \[
      \binom{8+4-1}{4-1}=\binom{11}{3}=165.
      \]
      
      All solutions satisfy the digit limits.
      
      Final answer: 165

      552 tokens in, 145 out (42 of them reasoning) · 1.7 s · $0.0006 · 1 message on Pro · answered by openai/gpt-6.1-terra via OpenAI ·

    • DeepSeek V4.1 Flash

      Passed: Final answer 165: right.

      Let the digits be a, b, c, d with a ≥ 1 (it's a four-digit number) and b, c, d ≥ 0, and a+b+c+d = 9.
      
      Substitute a = a' + 1 where a' ≥ 0:
      a' + b + c + d = 8
      
      Number of non-negative solutions = C(8+3, 3) = C(11, 3) = 165.
      
      Final answer: 165

      417 tokens in, 197 out (92 of them reasoning) · 0.7 s · $0.0003 · 1 message on Pro · answered by deepseek/deepseek-v4.1-flash via Together ·

  5. 5. The highest of three dice is a 5

    Hard. GPT: 5 of 5 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-6.1 Terra and DeepSeek V4.1 Flash both passed this one. DeepSeek V4.1 Flash answered sooner (1.0 s to 1.9 s), and DeepSeek V4.1 Flash's reply cost less ($0.0002 to $0.0006); GPT-6.1 Terra wrote 90 tokens of reply, DeepSeek V4.1 Flash 81.

    • GPT-6.1 Terra

      Passed: Final answer 61/216: right.

      There are \(6^3 = 216\) possible outcomes. Of these, \(5^3\) have all dice at most 5, and \(4^3\) have all dice at most 4. So the probability is
      
      \[
      \frac{5^3-4^3}{6^3}=\frac{61}{216}.
      \]
      
      Final answer: 61/216

      567 tokens in, 112 out (22 of them reasoning) · 1.9 s · $0.0006 · 1 message on Pro · answered by openai/gpt-6.1-terra via OpenAI ·

    • DeepSeek V4.1 Flash

      Passed: Final answer 61/216: right.

      Count outcomes where all three dice are ≤ 5, then subtract those where all are ≤ 4 (so no 5 appears):
      
      - All ≤ 5: 5³ = 125
      - All ≤ 4: 4³ = 64
      
      Favorable outcomes: 125 − 64 = 61, out of 216 total.
      
      Final answer: 61/216

      433 tokens in, 150 out (69 of them reasoning) · 1.0 s · $0.0002 · 1 message on Pro · answered by deepseek/deepseek-v4.1-flash via Together ·

Each GPT model against each DeepSeek model

Every GPT model against every DeepSeek model on the same 5 math prompts: passes, the median wait and what the replies cost. A lead under 10% counts as close.

  • GPT-6 Astra vs DeepSeek V4 Pro: 5 of 5 each; they took about as long and DeepSeek V4 Pro cost 9.7× less.

  • GPT-6 Astra vs DeepSeek V4.1 Flash: 5 of 5 each; DeepSeek V4.1 Flash answered 2.8× sooner at the median and cost 34.2× less.

  • GPT-6.1 Sol vs DeepSeek V4 Pro: 5 of 5 each; GPT-6.1 Sol answered 1.5× sooner at the median and cost about the same.

  • GPT-6.1 Sol vs DeepSeek V4.1 Flash: 5 of 5 each; DeepSeek V4.1 Flash answered 1.9× sooner at the median and cost 3.4× less.

  • GPT-6 Sol vs DeepSeek V4 Pro: 5 of 5 each; GPT-6 Sol answered 1.2× 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: 5 of 5 each; DeepSeek V4.1 Flash answered 2.3× sooner at the median and cost 7.2× less.

  • GPT-6.1 Terra vs DeepSeek V4 Pro: 5 of 5 each; GPT-6.1 Terra answered 1.4× sooner at the median and cost 1.7× less.

  • GPT-6.1 Terra vs DeepSeek V4.1 Flash: 5 of 5 each; DeepSeek V4.1 Flash answered 2.0× sooner at the median and cost 2.1× less.

  • GPT-6 Luna vs DeepSeek V4 Pro: GPT-6 Luna 4 of 5, DeepSeek V4 Pro 5; GPT-6 Luna answered 1.2× sooner at the median and cost 8.9× less. GPT-6 Luna vs DeepSeek V4 Pro, on every job.

  • GPT-6 Luna vs DeepSeek V4.1 Flash: GPT-6 Luna 4 of 5, DeepSeek V4.1 Flash 5; DeepSeek V4.1 Flash answered 2.4× sooner at the median and GPT-6 Luna cost 2.5× less. GPT-6 Luna vs DeepSeek V4.1 Flash, on every job.

How the math replies are scored

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 math prompts are scored, and each one in full.

The differences at a glance

What follows from each model's facts in our catalog.

  • The lineups

    GPT: 5 models, GPT-6 Astra, GPT-6.1 Sol, GPT-6 Sol, GPT-6.1 Terra, 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.

  • Plans

    Every plan has messages every month on them, GPT-6 Astra from Pro up. Every plan is month to month: cancel in two clicks.

Every GPT and DeepSeek model's context window, files and API price: DeepSeek vs ChatGPT.

Where your messages go

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.

Bar chart: Prompts passed in our test runs, math. DeepSeek V4.1 Flash: 5 of 5; GPT-6.1 Terra: 5 of 5; DeepSeek V4 Pro: 5 of 5; GPT-6.1 Sol: 5 of 5; GPT-6 Sol: 5 of 5; GPT-6 Astra: 5 of 5; GPT-6 Luna: 4 of 5.
Our test runs of September 27, 28, 29 and October 2, 7, 8, 9, 15, 27, 2026: the same prompts for every model, each reply checked the same way.

Questions

Which is better, ChatGPT or DeepSeek for math?

In our math test runs on October 27, 2026, GPT's 5 models passed 24 of 25; GPT-6 Astra, GPT-6.1 Sol and 2 more each passed 5 of 5. DeepSeek's 2 models passed 10 of 10; DeepSeek V4 Pro and DeepSeek V4.1 Flash each passed 5 of 5. GPT-6.1 Terra and DeepSeek V4.1 Flash each passed 5 of the 5 prompts, so these math prompts don't split GPT and DeepSeek; the replies on this page show how they differ.

Is GPT or DeepSeek cheaper?

In llmwise, the least expensive GPT model is GPT-6 Luna (60 messages a day on Pro), and the least expensive DeepSeek model is DeepSeek V4.1 Flash (60 messages a day on Pro). At API list prices (October 2026), a typical message of 4,000 tokens in and 700 out costs $0.0008 on GPT-6 Luna and $0.0020 on DeepSeek V4.1 Flash.

Can I use GPT and DeepSeek in the same chat?

Yes. Ask GPT-6.1 Terra a question, then switch the picker to DeepSeek V4.1 Flash and ask again: DeepSeek V4.1 Flash sees the whole conversation, GPT-6.1 Terra's answer included.

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.