Claude vs DeepSeek for math

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.

Math: Claude Haiku 4.5 passed 5 of 5 and DeepSeek V4.1 Flash 5 of 5. DeepSeek V4.1 Flash answered 1.3× sooner at the median, 2.2 s against 1.7 s. DeepSeek V4.1 Flash cost 6.6× less, $0.0077 against $0.0012 for the 5 replies. Claude Haiku 4.5's replies ran 125% longer, in tokens of reply, thinking not counted.

1. 1. A discount, then sales tax

   Everyday. Claude: 5 of 5 models passed, Claude Haiku 4.5 the cheapest at $0.0007 · DeepSeek: 2 of 2 models passed, DeepSeek V4 Pro the cheapest at $0.0002. The prompt in full.

   Claude Haiku 4.5 and DeepSeek V4.1 Flash both passed this one. DeepSeek V4.1 Flash answered sooner (0.6 s to 1.1 s), and DeepSeek V4.1 Flash's reply cost less ($0.0002 to $0.0007); Claude Haiku 4.5 wrote 63 tokens of reply, DeepSeek V4.1 Flash 41.

   • Claude Haiku 4.5

     Passed: Final answer 66: right.

     **Working:**
     
     1. Original price: $80
     2. Apply 25% discount: $80 × 0.75 = $60
     3. Add 10% sales tax: $60 × 1.10 = $66
     
     Final answer: $66

     429 tokens in, 63 out · 1.1 s · $0.0007 · 1 message on Pro · answered by anthropic/claude-haiku-4.5 via Amazon Bedrock · September 27, 2026

   • DeepSeek V4.1 Flash

     Passed: Final answer 66: right.

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

     440 tokens in, 64 out (23 of them reasoning) · 0.6 s · $0.0002 · 1 message on Pro · answered by deepseek/deepseek-v4.1-flash via Parasail · September 27, 2026

2. 2. Pens at 3 for $4

   Everyday. Claude: 5 of 5 models passed, Claude Haiku 4.5 the cheapest at $0.0014 · DeepSeek: 2 of 2 models passed, DeepSeek V4.1 Flash the cheapest at $0.0003. The prompt in full.

   Claude Haiku 4.5 and DeepSeek V4.1 Flash both passed this one. Claude Haiku 4.5 answered sooner (2.2 s to 25.6 s), and DeepSeek V4.1 Flash's reply cost less ($0.0003 to $0.0014); Claude Haiku 4.5 wrote 199 tokens of reply, DeepSeek V4.1 Flash 138.

   • Claude Haiku 4.5

     Passed: Final answer Buy 3 packs of 3 pens and 1 single pen for 13.50: right.

     # Pen Purchase Problem
     
     **Option 1: Buy 3-packs only**
     - 10 pens ÷ 3 = 3 packs with 1 pen left over
     - 3 packs × $4 = $12
     - 1 single pen × $1.50 = $1.50
     - Total: $13.50
     
     **Option 2: Buy all single pens**
     - 10 pens × $1.50 = $15.00
     
     **Option 3: Buy 3 packs + adjust**
     - Same as Option 1: $13.50
     
     The cheapest way is to buy three 3-packs ($12) and one single pen ($1.50).
     
     Final answer: Buy 3 packs of 3 pens and 1 single pen for $13.50

     434 tokens in, 199 out · 2.2 s · $0.0014 · 1 message on Pro · answered by anthropic/claude-haiku-4.5 via Amazon Bedrock · September 27, 2026

   • DeepSeek V4.1 Flash

     Passed: Final answer Buy three 3-packs (9 pens) plus one single pen for 13.50.: right.

     **Working:**
     - 3-pack price per pen: $4 ÷ 3 ≈ $1.33 (best value)
     - Buy as many 3-packs as fit: 3 packs = 9 pens = $12, plus 1 single at $1.50 = **$13.50**
     - Check other combinations: 2 packs + 4 singles = $8 + $6 = $14; 1 pack + 7 singles = $14.50; 10 singles = $15
     - The $13.50 option is cheapest.
     
     Final answer: Buy three 3-packs (9 pens) plus one single pen for $13.50.

     442 tokens in, 854 out (716 of them reasoning) · 25.6 s · $0.0003 · 1 message on Pro · answered by deepseek/deepseek-v4.1-flash via Morph · September 27, 2026

3. 3. Compound interest over three years

   Everyday. Claude: 5 of 5 models passed, Claude Haiku 4.5 the cheapest at $0.0013 · DeepSeek: 2 of 2 models passed, DeepSeek V4.1 Flash the cheapest at $0.0002. The prompt in full.

   Claude Haiku 4.5 and DeepSeek V4.1 Flash both passed this one. DeepSeek V4.1 Flash answered sooner (1.5 s to 3.5 s), and DeepSeek V4.1 Flash's reply cost less ($0.0002 to $0.0013); Claude Haiku 4.5 wrote 173 tokens of reply, DeepSeek V4.1 Flash 104.

   • Claude Haiku 4.5

     Passed: Final answer 315.25: right.

     I'll calculate the compound interest for you.
     
     **Formula:** A = P(1 + r)^n
     
     Where:
     - P = Principal = $2,000
     - r = Annual interest rate = 0.05
     - n = Number of years = 3
     
     **Calculation:**
     - Year 1: $2,000 × 1.05 = $2,100.00
     - Year 2: $2,100 × 1.05 = $2,205.00
     - Year 3: $2,205 × 1.05 = $2,315.25
     
     **Interest earned:**
     $2,315.25 - $2,000.00 = $315.25
     
     Final answer: $315.25

     430 tokens in, 173 out · 3.5 s · $0.0013 · 1 message on Pro · answered by anthropic/claude-haiku-4.5 via Amazon Bedrock · September 27, 2026

   • DeepSeek V4.1 Flash

     Passed: Final answer 315.25: right.

     **Working:**
     
     - Year 1: $2,000 × 1.05 = $2,100.00
     - Year 2: $2,100.00 × 1.05 = $2,205.00
     - Year 3: $2,205.00 × 1.05 = $2,315.25
     
     Interest earned = $2,315.25 − $2,000.00 = **$315.25**
     
     Final answer: $315.25

     440 tokens in, 186 out (82 of them reasoning) · 1.5 s · $0.0002 · 1 message on Pro · answered by deepseek/deepseek-v4.1-flash via CoreWeave · September 27, 2026

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

   Hard. Claude: 5 of 5 models passed, Claude Haiku 4.5 the cheapest at $0.0024 · DeepSeek: 2 of 2 models passed, DeepSeek V4.1 Flash the cheapest at $0.0002. The prompt in full.

   Claude Haiku 4.5 and DeepSeek V4.1 Flash both passed this one. DeepSeek V4.1 Flash answered sooner (1.7 s to 2.9 s), and DeepSeek V4.1 Flash's reply cost less ($0.0002 to $0.0024); Claude Haiku 4.5 wrote 404 tokens of reply, DeepSeek V4.1 Flash 112.

   • Claude Haiku 4.5

     Passed: Final answer 165: right.

     I need to find four-digit positive integers where the digits sum to 9.
     
     A four-digit positive integer has the form $\overline{abcd}$ where:
     - $a \in \{1,2,...,9\}$ (first digit can't be 0)
     - $b,c,d \in \{0,1,...,9\}$
     - $a + b + c + d = 9$
     
     **Method: Change of variables**
     
     Let $a' = a - 1$, so $a' \in \{0,1,...,8\}$.
     
     Then: $a' + 1 + b + c + d = 9$
     
     This gives: $a' + b + c + d = 8$
     
     Now I need to count non-negative integer solutions where:
     - $a' \in \{0,1,...,8\}$
     - $b,c,d \in \{0,1,...,9\}$
     - $a' + b + c + d = 8$
     
     Since the sum is only 8, each variable is automatically ≤ 8 (and thus $a' \leq 8$ and $b,c,d \leq 9$ are automatically satisfied).
     
     This is a standard stars and bars problem: distribute 8 identical objects into 4 distinct bins with no restrictions.
     
     The number of solutions is:
     $$\binom{8 + 4 - 1}{4 - 1} = \binom{11}{3} = \frac{11 \times 10 \times 9}{3 \times 2 \times 1} = \frac{990}{6} = 165$$
     
     Final answer: 165

     404 tokens in, 404 out · 2.9 s · $0.0024 · 1 message on Pro · answered by anthropic/claude-haiku-4.5 via Amazon Bedrock · September 27, 2026

   • DeepSeek V4.1 Flash

     Passed: Final answer 165: right.

     Let the digits be $a,b,c,d$ with $a \ge 1$ (four-digit) and $b,c,d \ge 0$, and $a+b+c+d=9$.
     
     Set $a' = a-1 \ge 0$. Then $a'+b+c+d = 8$.
     
     Number of nonnegative solutions to a sum of 8 across 4 variables:
     
     $$\binom{8+4-1}{4-1} = \binom{11}{3} = 165$$
     
     Final answer: 165

     417 tokens in, 216 out (104 of them reasoning) · 1.7 s · $0.0002 · 1 message on Pro · answered by deepseek/deepseek-v4.1-flash via Novita · September 27, 2026

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

   Hard. Claude: 5 of 5 models passed, Claude Haiku 4.5 the cheapest at $0.0018 · DeepSeek: 2 of 2 models passed, DeepSeek V4.1 Flash the cheapest at $0.0002. The prompt in full.

   Claude Haiku 4.5 and DeepSeek V4.1 Flash both passed this one. Claude Haiku 4.5 answered sooner (2.2 s to 2.5 s), and DeepSeek V4.1 Flash's reply cost less ($0.0002 to $0.0018); Claude Haiku 4.5 wrote 276 tokens of reply, DeepSeek V4.1 Flash 100.

   • Claude Haiku 4.5

     Passed: Final answer 61/216: right.

     I need to find the probability that the maximum of three dice is exactly 5.
     
     **Working:**
     
     For the maximum to be exactly 5:
     - At least one die shows 5
     - No die shows 6
     
     **Count favorable outcomes:**
     - All three dice must show values from {1, 2, 3, 4, 5}
     - At least one die must show 5
     
     Using complementary counting:
     - Total ways to roll three dice with values in {1,2,3,4,5}: 5³ = 125
     - Ways with no 5 (all from {1,2,3,4}): 4³ = 64
     - Ways with at least one 5: 125 - 64 = 61
     
     **Total possible outcomes:** 6³ = 216
     
     **Probability:** 61/216
     
     Let me verify this is in lowest terms by checking if 61 and 216 share factors:
     - 61 is prime
     - 216 = 2³ × 3³
     - Since 61 is prime and doesn't equal 2 or 3, gcd(61, 216) = 1
     
     Final answer: 61/216

     420 tokens in, 276 out · 2.2 s · $0.0018 · 1 message on Pro · answered by anthropic/claude-haiku-4.5 via Amazon Bedrock · September 27, 2026

   • DeepSeek V4.1 Flash

     Passed: Final answer 61/216: right.

     Total outcomes: 6³ = 216.
     
     Max exactly 5 means all dice show ≤ 5, but not all show ≤ 4:
     
     - All ≤ 5: 5³ = 125
     - All ≤ 4: 4³ = 64
     
     So favorable outcomes: 125 − 64 = 61.
     
     Probability = 61/216 (already in lowest terms, since 61 is prime and doesn't divide 216).
     
     Final answer: 61/216

     432 tokens in, 155 out (55 of them reasoning) · 2.5 s · $0.0002 · 1 message on Pro · answered by deepseek/deepseek-v4.1-flash via CoreWeave · September 27, 2026

• Claude Fable 5.1 vs DeepSeek V4 Pro: 5 of 5 each; DeepSeek V4 Pro answered 1.7× sooner at the median and cost 17.6× less.

• Claude Fable 5.1 vs DeepSeek V4.1 Flash: 5 of 5 each; DeepSeek V4.1 Flash answered 2.7× sooner at the median and cost 50.1× less.

• Claude Opus 5.5 vs DeepSeek V4 Pro: 5 of 5 each; DeepSeek V4 Pro answered 1.4× sooner at the median and cost 9.0× less. Claude Opus 5.5 vs DeepSeek V4 Pro, on every job.

• Claude Opus 5.5 vs DeepSeek V4.1 Flash: 5 of 5 each; DeepSeek V4.1 Flash answered 2.3× sooner at the median and cost 25.7× less.

• Claude Sonnet 5.5 vs DeepSeek V4 Pro: 5 of 5 each; Claude Sonnet 5.5 answered 1.6× sooner at the median and DeepSeek V4 Pro cost 3.7× less.

• Claude Sonnet 5.5 vs DeepSeek V4.1 Flash: 5 of 5 each; they took about as long and DeepSeek V4.1 Flash cost 10.6× less.

• Claude Sonnet 5 vs DeepSeek V4 Pro: 5 of 5 each; DeepSeek V4 Pro answered 1.4× sooner at the median and cost 5.0× less. Claude Sonnet 5 vs DeepSeek V4 Pro, on every job.

• Claude Sonnet 5 vs DeepSeek V4.1 Flash: 5 of 5 each; DeepSeek V4.1 Flash answered 2.3× sooner at the median and cost 14.2× less.

• Claude Haiku 4.5 vs DeepSeek V4 Pro: 5 of 5 each; Claude Haiku 4.5 answered 1.2× sooner at the median and DeepSeek V4 Pro cost 2.3× less. Claude Haiku 4.5 vs DeepSeek V4 Pro, on every job.

• Claude Haiku 4.5 vs DeepSeek V4.1 Flash: 5 of 5 each; DeepSeek V4.1 Flash answered 1.3× sooner at the median and cost 6.6× less. Claude Haiku 4.5 vs DeepSeek V4.1 Flash, on every job.

Every Claude 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 lineups

  Claude: 5 models, Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5, Claude Sonnet 5, and Claude Haiku 4.5. DeepSeek: 2 models, DeepSeek V4 Pro and DeepSeek V4.1 Flash.

• Price per message

  The least expensive Claude model is Claude Haiku 4.5 (250 messages a month on Pro); the least expensive DeepSeek model is DeepSeek V4.1 Flash (60 messages a day on Pro).

• Context window

  Claude goes up to 1M tokens (Claude Fable 5.1); 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

  Free's one-time trial of 5 messages covers Claude Sonnet 5.5, Claude Sonnet 5, Claude Haiku 4.5, DeepSeek V4 Pro, and DeepSeek V4.1 Flash. Paid plans have every model, with messages every month.

In llmwise, a message to Claude goes to its maker, Anthropic, 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.