Model vs model
DeepSeek V4 Pro vs Kimi K3
DeepSeek V4 Pro and Kimi K3 are both in llmwise. What each gets on every plan, what it reads, how it's served and what happens when its provider fails, from the catalog and the code that runs them.
Model prices and specs checked against OpenRouter's DeepSeek V4 Pro page, OpenRouter's Kimi K3 page. Updated .
Short answer
DeepSeek V4 Pro gets twice as many messages: up to 250 messages a month on Pro, against up to 125 messages a month for Kimi K3. Otherwise, only Kimi K3 reads images. In our test runs, DeepSeek V4 Pro passed 49 of the 50 prompts both answered and Kimi K3 46; 5 prompts split them, most on summarization (5 to 3).
DeepSeek V4 Pro vs Kimi K3, prompt by prompt
Every prompt DeepSeek V4 Pro and Kimi K3 both answered, compared directly, their biggest differences first. One run each, through OpenRouter: a wait depends on the provider and the load that day, so a lead under 10% counts as close.
Of the 50 prompts both answered, both passed 45, only DeepSeek V4 Pro passed 4, only Kimi K3 passed 1, and neither passed 0. DeepSeek V4 Pro answered sooner on 25 of the 50 and Kimi K3 on 21; the rest were within 10% of each other. The 50 replies cost $0.1827 on DeepSeek V4 Pro and $0.1938 on Kimi K3: about the same for both.
The 5 prompts only one of DeepSeek V4 Pro and Kimi K3 passed
A product announcement with five rules (writing): DeepSeek V4 Pro passed and Kimi K3 didn't. DeepSeek V4 Pro: Graded 4.3 of 5 on average (lowest 4). Kimi K3: Graded 4.3 of 5 on average (lowest 4); but doesn't end with a question.
An article in three bullets (summarization): DeepSeek V4 Pro passed and Kimi K3 didn't. DeepSeek V4 Pro: Graded 4.3 of 5 on average (lowest 4). Kimi K3: Graded 4.0 of 5 on average (lowest 3); but 62 words, over the 60 allowed.
An email thread in one sentence (summarization): DeepSeek V4 Pro passed and Kimi K3 didn't. DeepSeek V4 Pro: Graded 4.7 of 5 on average (lowest 4). Kimi K3: Graded 5.0 of 5 on average (lowest 5); but 32 words, over the 30 allowed.
A frustrated customer (customer support): Kimi K3 passed and DeepSeek V4 Pro didn't. DeepSeek V4 Pro: Graded 2.3 of 5 on average (lowest 1). Kimi K3: Graded 4.0 of 5 on average (lowest 3).
A refund request outside the window (customer support): DeepSeek V4 Pro passed and Kimi K3 didn't. DeepSeek V4 Pro: Graded 4.7 of 5 on average (lowest 4). Kimi K3: Graded 4.0 of 5 on average (lowest 2).
Job by job, the widest gaps first
Summarization: DeepSeek V4 Pro passed 5 of 5 and Kimi K3 3 of 5. DeepSeek V4 Pro answered 1.6× sooner at the median, 2.1 s against 3.4 s. DeepSeek V4 Pro cost 3.8× less, $0.0046 against $0.0172 for the 5 replies. Kimi K3's replies ran 27% longer, in tokens of reply, thinking not counted.
Writing: DeepSeek V4 Pro passed 5 of 5 and Kimi K3 4 of 5. Kimi K3 answered 1.3× sooner at the median, 3.9 s against 3.0 s. DeepSeek V4 Pro cost 3.6× less, $0.0054 against $0.0192 for the 5 replies. Kimi K3's replies ran 41% longer, in tokens of reply, thinking not counted.
Math: DeepSeek V4 Pro passed 5 of 5 and Kimi K3 5 of 5. DeepSeek V4 Pro answered 2.0× sooner at the median, 2.7 s against 5.4 s. DeepSeek V4 Pro cost 4.3× less, $0.0044 against $0.0187 for the 5 replies. Their replies ran to about the same length.
Agents and tool use: DeepSeek V4 Pro passed 5 of 5 and Kimi K3 5 of 5. DeepSeek V4 Pro answered 1.2× sooner at the median, 1.6 s against 2.0 s. DeepSeek V4 Pro cost 4.1× less, $0.0033 against $0.0135 for the 5 replies. Kimi K3's replies ran 12% longer, in tokens of reply, thinking not counted.
Coding: DeepSeek V4 Pro passed 5 of 5 and Kimi K3 5 of 5. Kimi K3 answered 28.9× sooner at the median, 119.6 s against 4.1 s. Kimi K3 cost 3.8× less, $0.1200 against $0.0316 for the 5 replies. Kimi K3's replies ran 13% longer, in tokens of reply, thinking not counted.
Customer support: DeepSeek V4 Pro passed 4 of 5 and Kimi K3 4 of 5. Their median waits were close, 4.1 s against 4.3 s. DeepSeek V4 Pro cost 3.0× less, $0.0077 against $0.0228 for the 5 replies. Kimi K3's replies ran 41% longer, in tokens of reply, thinking not counted.
RAG and answering from documents: DeepSeek V4 Pro passed 5 of 5 and Kimi K3 5 of 5. DeepSeek V4 Pro answered 1.3× sooner at the median, 2.1 s against 2.6 s. DeepSeek V4 Pro cost 2.4× less, $0.0030 against $0.0073 for the 5 replies. Kimi K3's replies ran 65% longer, in tokens of reply, thinking not counted.
Data analysis: DeepSeek V4 Pro passed 5 of 5 and Kimi K3 5 of 5. Kimi K3 answered 1.2× sooner at the median, 4.1 s against 3.5 s. DeepSeek V4 Pro cost 2.4× less, $0.0143 against $0.0339 for the 5 replies. Kimi K3's replies ran 38% longer, in tokens of reply, thinking not counted.
Translation: DeepSeek V4 Pro passed 5 of 5 and Kimi K3 5 of 5. DeepSeek V4 Pro answered 1.2× sooner at the median, 4.3 s against 5.2 s. DeepSeek V4 Pro cost 1.8× less, $0.0100 against $0.0179 for the 5 replies. Kimi K3's replies ran 88% longer, in tokens of reply, thinking not counted.
SQL: DeepSeek V4 Pro passed 5 of 5 and Kimi K3 5 of 5. Kimi K3 answered 5.1× sooner at the median, 8.0 s against 1.6 s. DeepSeek V4 Pro cost 1.2× less, $0.0101 against $0.0118 for the 5 replies. Their replies ran to about the same length.
All 50 prompts: who passed, who answered sooner, who cost less
| Prompt | Result | Sooner | Cheaper |
|---|---|---|---|
| Turn a title into a URL slug | Both passed | Kimi K3, 1.5×took 3.9 s and 2.6 s | DeepSeek V4 Pro, 2.4×cost $0.0015 and $0.0035 |
| Parse a duration like “1h 30m” | Both passed | Kimi K3, 16.8×took 119.6 s and 7.1 s | Kimi K3, 5.9×cost $0.0366 and $0.0063 |
| Merge overlapping intervals | Both passed | Kimi K3, 3.3×took 6.8 s and 2.0 s | DeepSeek V4 Pro, 1.7×cost $0.0021 and $0.0036 |
| Evaluate an arithmetic expression, no eval | Both passed | Kimi K3, 14.7×took 145.7 s and 9.9 s | Kimi K3, 3.9×cost $0.0440 and $0.0113 |
| Parse CSV with quoted fields | Both passed | Kimi K3, 29.6×took 122.3 s and 4.1 s | Kimi K3, 5.2×cost $0.0357 and $0.0069 |
| Announce a second bakery shop on LinkedIn | Both passed | Closetook 3.9 s and 3.9 s | DeepSeek V4 Pro, 4.8×cost $0.0010 and $0.0046 |
| Rewrite corporate jargon in plain words | Both passed | Kimi K3, 2.9×took 5.8 s and 2.0 s | DeepSeek V4 Pro, 1.2×cost $0.0024 and $0.0030 |
| Decline a meeting and offer two times | Both passed | Kimi K3, 1.1×took 3.3 s and 3.0 s | DeepSeek V4 Pro, 5.7×cost $0.0006 and $0.0034 |
| A product announcement with five rules | Only DeepSeek V4 Pro | DeepSeek V4 Pro, 1.3×took 1.8 s and 2.3 s | DeepSeek V4 Pro, 6.6×cost $0.0005 and $0.0033 |
| Argue both sides of free buses | Both passed | Kimi K3, 1.1×took 4.1 s and 3.6 s | DeepSeek V4 Pro, 5.5×cost $0.0009 and $0.0049 |
| A discount, then sales tax | Both passed | DeepSeek V4 Pro, 2.8×took 1.8 s and 5.0 s | DeepSeek V4 Pro, 7.3×cost $0.0005 and $0.0037 |
| Pens at 3 for $4 | Both passed | DeepSeek V4 Pro, 2.0×took 2.7 s and 5.4 s | DeepSeek V4 Pro, 4.4×cost $0.0010 and $0.0046 |
| Compound interest over three years | Both passed | DeepSeek V4 Pro, 1.7×took 3.4 s and 5.9 s | DeepSeek V4 Pro, 5.9×cost $0.0008 and $0.0044 |
| Four-digit numbers whose digits sum to 9 | Both passed | DeepSeek V4 Pro, 1.5×took 4.0 s and 6.0 s | DeepSeek V4 Pro, 2.7×cost $0.0012 and $0.0032 |
| The highest of three dice is a 5 | Both passed | Closetook 2.5 s and 2.4 s | DeepSeek V4 Pro, 3.0×cost $0.0009 and $0.0027 |
| An article in three bullets | Only DeepSeek V4 Pro | DeepSeek V4 Pro, 1.6×took 1.8 s and 2.8 s | DeepSeek V4 Pro, 2.8×cost $0.0010 and $0.0028 |
| An email thread in one sentence | Only DeepSeek V4 Pro | DeepSeek V4 Pro, 5.2×took 1.2 s and 6.4 s | DeepSeek V4 Pro, 5.8×cost $0.0005 and $0.0030 |
| Decisions and action items from a meeting | Both passed | Kimi K3, 1.6×took 5.4 s and 3.4 s | DeepSeek V4 Pro, 4.0×cost $0.0011 and $0.0044 |
| A quarterly memo for the CEO | Both passed | DeepSeek V4 Pro, 6.8×took 2.3 s and 15.7 s | DeepSeek V4 Pro, 3.5×cost $0.0014 and $0.0049 |
| A study with a negative result | Both passed | DeepSeek V4 Pro, 1.6×took 2.1 s and 3.2 s | DeepSeek V4 Pro, 3.9×cost $0.0005 and $0.0021 |
| The region with the most revenue | Both passed | Kimi K3, 1.8×took 4.1 s and 2.2 s | DeepSeek V4 Pro, 2.5×cost $0.0020 and $0.0049 |
| Average order value in August | Both passed | DeepSeek V4 Pro, 1.2×took 2.2 s and 2.7 s | DeepSeek V4 Pro, 4.5×cost $0.0012 and $0.0054 |
| Revenue change from July to August | Both passed | DeepSeek V4 Pro, 1.2×took 5.2 s and 6.4 s | DeepSeek V4 Pro, 3.2×cost $0.0020 and $0.0064 |
| A median, filtered two ways | Both passed | DeepSeek V4 Pro, 3.0×took 1.2 s and 3.5 s | DeepSeek V4 Pro, 4.0×cost $0.0011 and $0.0042 |
| Correlation between ad spend and sign-ups | Both passed | DeepSeek V4 Pro, 2.5×took 8.1 s and 20.2 s | DeepSeek V4 Pro, 1.6×cost $0.0081 and $0.0131 |
| A late order | Both passed | Kimi K3, 1.6×took 4.1 s and 2.5 s | DeepSeek V4 Pro, 2.0×cost $0.0023 and $0.0045 |
| A return inside the window | Both passed | DeepSeek V4 Pro, 3.2×took 1.3 s and 4.3 s | DeepSeek V4 Pro, 5.0×cost $0.0007 and $0.0033 |
| A frustrated customer | Only Kimi K3 | Kimi K3, 1.7×took 9.5 s and 5.6 s | DeepSeek V4 Pro, 3.5×cost $0.0018 and $0.0063 |
| A refund request outside the window | Only DeepSeek V4 Pro | Closetook 2.5 s and 2.6 s | DeepSeek V4 Pro, 3.7×cost $0.0012 and $0.0045 |
| A message with a planted instruction | Both passed | DeepSeek V4 Pro, 1.7×took 6.7 s and 11.5 s | DeepSeek V4 Pro, 2.4×cost $0.0017 and $0.0041 |
| A delivery message into Spanish | Both passed | Kimi K3, 3.0×took 8.8 s and 2.9 s | Kimi K3, 1.4×cost $0.0044 and $0.0031 |
| A product description into French | Both passed | DeepSeek V4 Pro, 4.0×took 1.3 s and 5.4 s | DeepSeek V4 Pro, 5.8×cost $0.0006 and $0.0037 |
| A meeting note into German | Both passed | Kimi K3, 1.4×took 4.3 s and 3.0 s | Closecost $0.0015 and $0.0014 |
| Idioms into natural Japanese | Both passed | DeepSeek V4 Pro, 1.9×took 2.7 s and 5.2 s | DeepSeek V4 Pro, 4.8×cost $0.0011 and $0.0053 |
| A lease clause into Brazilian Portuguese | Both passed | Kimi K3, 1.6×took 16.4 s and 10.1 s | DeepSeek V4 Pro, 1.8×cost $0.0024 and $0.0044 |
| Customers in one country | Both passed | Closetook 1.2 s and 1.1 s | DeepSeek V4 Pro, 4.5×cost $0.0004 and $0.0018 |
| Count orders by status | Both passed | DeepSeek V4 Pro, 1.4×took 1.1 s and 1.5 s | DeepSeek V4 Pro, 2.1×cost $0.0004 and $0.0009 |
| Revenue by category | Both passed | Kimi K3, 5.1×took 8.0 s and 1.6 s | Kimi K3, 1.3×cost $0.0025 and $0.0019 |
| Every customer, even those without orders | Both passed | Kimi K3, 4.1×took 10.4 s and 2.5 s | Closecost $0.0032 and $0.0031 |
| Monthly revenue with a running total | Both passed | Kimi K3, 6.6×took 16.6 s and 2.5 s | DeepSeek V4 Pro, 1.2×cost $0.0036 and $0.0042 |
| A fact from one section | Both passed | DeepSeek V4 Pro, 6.1×took 1.1 s and 6.5 s | DeepSeek V4 Pro, 2.0×cost $0.0005 and $0.0010 |
| Core hours and start times | Both passed | DeepSeek V4 Pro, 1.5×took 0.8 s and 1.2 s | DeepSeek V4 Pro, 3.1×cost $0.0004 and $0.0011 |
| Two sections in one answer | Both passed | DeepSeek V4 Pro, 1.3×took 2.1 s and 2.6 s | DeepSeek V4 Pro, 2.6×cost $0.0008 and $0.0020 |
| A later amendment changes the answer | Both passed | DeepSeek V4 Pro, 1.9×took 2.5 s and 4.8 s | DeepSeek V4 Pro, 2.5×cost $0.0009 and $0.0022 |
| A question the handbook doesn't answer | Both passed | Kimi K3, 9.4×took 8.8 s and 0.9 s | DeepSeek V4 Pro, 1.9×cost $0.0005 and $0.0010 |
| Pick the tool and work out the date | Both passed | DeepSeek V4 Pro, 2.3×took 1.6 s and 3.7 s | DeepSeek V4 Pro, 6.2×cost $0.0005 and $0.0032 |
| Convert a currency | Both passed | DeepSeek V4 Pro, 3.3×took 0.9 s and 2.9 s | DeepSeek V4 Pro, 6.1×cost $0.0004 and $0.0025 |
| Book a meeting from a sentence | Both passed | Kimi K3, 1.5×took 2.3 s and 1.6 s | DeepSeek V4 Pro, 3.7×cost $0.0008 and $0.0031 |
| Search, but don't book | Both passed | DeepSeek V4 Pro, 1.4×took 1.4 s and 2.0 s | DeepSeek V4 Pro, 5.0×cost $0.0007 and $0.0034 |
| Two calls with a unit conversion | Both passed | Kimi K3, 2.2×took 3.7 s and 1.7 s | DeepSeek V4 Pro, 1.5×cost $0.0008 and $0.0012 |
DeepSeek V4 Pro vs Kimi K3 in our test runs
DeepSeek V4 Pro and Kimi K3 on the same prompts, job by job: how many replies passed their check.
Based on 100 of our test runs on , through OpenRouter with the app's own prompt and settings. Updated .
| Job | DeepSeek V4 Pro | Kimi K3 |
|---|---|---|
| Coding | 5 of 5 | 5 of 5 |
| Writing | 5 of 5 | 4 of 5 |
| Math | 5 of 5 | 5 of 5 |
| Summarization | 5 of 5 | 3 of 5 |
| Data analysis | 5 of 5 | 5 of 5 |
| Customer support | 4 of 5 | 4 of 5 |
| Translation | 5 of 5 | 5 of 5 |
| SQL | 5 of 5 | 5 of 5 |
| RAG and answering from documents | 5 of 5 | 5 of 5 |
| Agents and tool use | 5 of 5 | 5 of 5 |
| All jobs | 49 of 50 · 1 message each · $0.0037 a reply | 46 of 50 · 1 message each · $0.0039 a reply |
The same prompts, sent to both through llmwise's own pipeline and checked the same way: five for each job. A bold count is the one that passed more. Messages: what each reply counted as on Pro. Cost: what OpenRouter charged us per reply, on average. Every prompt, and how it's scored.
Where they differed most, summarization: both replies to a prompt only one of them passed.
An everyday prompt: “An article in three bullets”, in full.
DeepSeek V4 Pro
Passed: Graded 4.3 of 5 on average (lowest 4).
- Bike trips on Market Street rose from 900 to 2,300 daily during the trial, while car traffic fell 14%. - Cyclist injuries dropped from nine to two, but shop owners were split: 27 saw more customers, 19 fewer. - The council votes in November; permanence would cost another €1.1 million.
534 tokens in, 196 out (124 of them reasoning) · 1.8 s · $0.0010 · 1 message on Pro · answered by deepseek/deepseek-v4-pro-0813 via Wafer ·
Kimi K3
Failed: Graded 4.0 of 5 on average (lowest 3); but 62 words, over the 60 allowed.
- Aldmoor's six-month protected bike lane trial on Market Street ended; the council votes in November on making them permanent. - Bike trips rose from 900 to 2,300 daily, cyclist injuries dropped from nine to two, while car traffic fell 14% with only slightly longer rush-hour drive times. - Businesses were split on customer impact; installation cost €410,000, with permanent lanes requiring €1.1 million more.
608 tokens in, 109 out (10 of them reasoning) · 2.8 s · $0.0028 · 1 message on Pro · answered by moonshotai/kimi-k3 via Phala ·
DeepSeek V4 Pro and Kimi K3 on every plan
Whether the one-time free trial reaches each model, then each paid plan's messages on it.
| Plan | Price | DeepSeek V4 Pro | Kimi K3 |
|---|---|---|---|
| Free | $0 | Not in the trial | Not in the trial |
| Plus | $10 a month | Up to 125 a month | Up to 62 a month |
| Pro | $20 a month | Up to 250 a month | Up to 125 a month |
| Max | $50 a month | Up to 800 a month | Up to 400 a month |
| Ultra | $100 a month | Up to 1,800 a month | Up to 900 a month |
| Studio | $200 a month | Up to 4,000 a month | Up to 2,000 a month |
Prices don't include tax, which is added where it applies and shown before you pay. A paid plan's month is one allowance shared by every model, so each monthly count is the most you get if all of it goes to that model. It renews each billing period; everyday models refill daily at 00:00 UTC. Long chats count more per reply. How pricing works.
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.
What differs
Messages on Pro
DeepSeek V4 Pro gets twice as many messages: up to 250 messages a month on Pro, against up to 125 messages a month for Kimi K3.
Context window
Both take up to 1.05M tokens of context. A chat in llmwise holds up to 200k tokens, which fits in either, so the difference shows only through each maker's own API.
Images and PDFs
DeepSeek V4 Pro doesn't read images. DeepSeek V4 Pro and Kimi K3 get a PDF's text rather than the file itself.
On Free
Neither is in the free trial: both need a paid plan, from Plus.
Where messages go
DeepSeek V4 Pro: Served through OpenRouter, only by hosts that don't store or train on prompts. The maker's own endpoint is never asked. Kimi K3: Served through OpenRouter, only by hosts that don't store or train on prompts. The maker's own endpoint is never asked.
Fact by fact
| Fact | DeepSeek V4 Pro | Kimi K3 |
|---|---|---|
| Context window | 1.05M tokens | 1.05M tokens |
| Reads images | No | Yes |
| PDFs | Text only | Text only |
| Reasoning | Yes | Yes |
| API price (October 2026) | $0.40 in / $4.00 out per million tokens | $3.00 in / $15.00 out per million tokens |
| A typical message at API prices (4,000 tokens in, 700 out) | $0.0044 | $0.0225 |
| A $10 top-up adds | 200 messages | 100 messages |
| Where a message goes | Served through OpenRouter, only by hosts that don't store or train on prompts. The maker's own endpoint is never asked. | Served through OpenRouter, only by hosts that don't store or train on prompts. The maker's own endpoint is never asked. |
| If the provider fails | When one host is down, OpenRouter moves the request to another host that meets the same rules. | When one host is down, OpenRouter moves the request to another host that meets the same rules. |
| Anthropic's safety fallback | Doesn't apply | Doesn't apply |
DeepSeek V4 Pro or Kimi K3?
From the facts above and our test runs: the rest is how their answers suit your work, which one chat can show you.
Pick DeepSeek V4 Pro: more messages for the money, up to 250 messages a month on Pro.
Pick Kimi K3: it reads images.
Each model's page, the families, and other pairs
DeepSeek V4 Pro vs Kimi K3 is one pair of models. The page below covers the whole families.
- DeepSeek V4 Pro: price, limits and messages on every plan
- Kimi K3: price, limits and messages on every plan
- DeepSeek vs Kimi
- Claude Opus 5.5 vs DeepSeek V4 Pro
- DeepSeek V4 Pro vs GLM 5.3
- Claude Sonnet 5 vs DeepSeek V4 Pro
- Claude Sonnet 5.5 vs Kimi K3
- Gemini 4 Argon vs Kimi K3
- Claude Fable 5.1 vs Kimi K3
- Every model-vs-model page
DeepSeek V4 Pro is a DeepSeek model; Kimi K3 is a Kimi model.
Questions
Is DeepSeek V4 Pro or Kimi K3 cheaper in llmwise?
DeepSeek V4 Pro gets twice as many messages: up to 250 messages a month on Pro, against up to 125 messages a month for Kimi K3. Every paid plan's monthly allowance is shared by the models on it, so each count is the most you get if it all goes to that model.
Can I try DeepSeek V4 Pro and Kimi K3 for free?
Neither is in the free trial of 5 messages: both need a paid plan, from Plus.
Which has the bigger context window, DeepSeek V4 Pro or Kimi K3?
Neither: both take 1.05M tokens. A chat in llmwise holds up to 200k tokens, which fits in either, so the difference shows only through each maker's own API.
Can I use DeepSeek V4 Pro and Kimi K3 in the same chat?
Yes. Pick DeepSeek V4 Pro for one message and Kimi K3 for the next; the second sees the whole chat, including the first one's answer.
Which did better in your test runs, DeepSeek V4 Pro or Kimi K3?
On the same 50 prompts, run on October 8, 2026, DeepSeek V4 Pro passed 49 and Kimi K3 passed 46. The table on this page has each job, and every reply is published.
Claude, GPT, Gemini, DeepSeek, Grok, Kimi, GLM, and Mistral, 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.