Compare / Kimi K3 vs Kimi K3.1 Pro

Kimi K3 vs Kimi K3.1 Pro

Which is cheaper, which is more capable, and which does better at each kind of work, from 64 results on 5 sources that measured both models.

Moonshot AIKimi K3 Released 16 Jul 2026 · $3.00 in · $15.00 out
Moonshot AIKimi K3.1 Pro Released 23 Oct 2026 · $1.75 in · $10.00 out
Kimi K3 wins 1513 too close to callKimi K3.1 Pro wins 36

Of the 64 results both models have, the better value on each. Results whose reported ranges overlap are too close to call.

Results as of 19 November 2026 (catalogue release 2026-11-19-3608133e8331).

Kimi K3 or Kimi K3.1 Pro? The short answer

Pick Kimi K3.1 Pro: it scores higher (Spring Prompt overall 67 against 46) at about the same price.

  • Spring Prompt overall: Kimi K3 46, Kimi K3.1 Pro 67 (out of 100; how the score works).
  • They cost about the same to run: $0.047 and $0.045 per product listing on CatalogBench.
  • Kimi K3.1 Pro is newer: released 23 Oct 2026, against 16 Jul 2026 for Kimi K3.
  • On our own benchmarks, on the results both models have: Kimi K3.1 Pro does better for marketing planning.
  • On third-party benchmarks, on the results both models have: Kimi K3.1 Pro does better for agents and tool use and reasoning and knowledge.

Head to head

46
Spring Prompt overall
67
$6.00
Price per million tokens, blendedlower is better
$3.81
1,048,576
Context window, tokens
1,048,576

Every fact side by side

Kimi K3Kimi K3.1 Pro
Price per million tokens$3.00 in · $15.00 out$1.75 in · $10.00 out
Blended (3 in : 1 out)$6.00$3.81
Cheapest host, blended$2.47 (Relace, FP4)$2.38 (Relace, FP4)
Cost per product listing (CatalogBench)$0.047$0.045
Cached input, per million$0.30 cached input–
Intelligence Index43.6–
Spring Prompt overall4667
Output speed, tokens a second41–
Context window1,048,576 tokens1,048,576 tokens
Released16 Jul 202623 Oct 2026 (newer)
WeightsOpenOpen
Developer based inChinaChina

Prices are the developer's list price where we have read it, otherwise the typical host on OpenRouter; see every model's price. Intelligence and speed from Artificial Analysis.

Which is better for what

Each model's rank among the models on sale today, for each kind of work, from our own benchmarks and licensed sources. Further right is better; the better rank is in green.

Product listingsTurning a sparse product feed and photos into listings that can go live 24 of 35 not measured
Decks from an analysisTurning a finished analysis into a deck you could present as it is 16 of 35 14 of 35
User surveysPlanning a user survey and reading its results without being misled not measured 23 of 26
Marketing planningPlanning a year of ad spend without overspending 20 of 35 5 of 35
Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls 40 of 159 1 of 159
Professional workReal tasks from banking, consulting and law, business documents and freelance projects 24 of 51 not measured
Reasoning and knowledgeHard questions across science, maths and general knowledge 18 of 187 9 of 187
WritingWhat people prefer in blind comparisons, and judged writing quality 12 of 169 not measured
Sticking to the factsSummarising without inventing things, and factual answers 27 of 139 not measured
Speed under pressureGood decisions against a real clock (fast chess) 36 of 38 29 of 38

Their slides, side by side

The same brief for both: Manchester store investment review for Hearthside Coffee, an invented company. First slide of each deck from DeckBench.

First slide of Kimi K3's deck for Hearthside Coffee
Kimi K3 · deck rating 1,194 · needs work
First slide of Kimi K3.1 Pro's deck for Hearthside Coffee
Kimi K3.1 Pro · deck rating 1,223 · needs work

Every slide, side by side →

From our research

Every shared result

Each model's best result on every measure both have, grouped by benchmark, with the setting that produced it. Under each measure, every model that benchmark has measured is a grey tick, best to the right, so you can see where the two sit in the field. The headline measures of each benchmark come first; the rest are under "Show all". Green marks the better value; where a source reports ranges that overlap, neither is marked.

MeasureKimi K3Kimi K3.1 Pro
BulletBench measured by Spring Prompt · 3 results
Cost per game · Bullet 60sUS dollars, lower is better 56 · 48 of 62 measured $0.0540low reasoning $0.0175low reasoning
Ladder Elo · Bullet 60sladder Elo, higher is better 59 · 54 of 62 measured 33low reasoning 147low reasoning
Games lost on time · Bullet 60s% of games, lower is better 61 · 57 of 62 measured 91.7%low reasoning 87.5%low reasoning
CatalogBench measured by Spring Prompt · 3 results
Cost per productUS dollars, lower is better 33 · 32 of 36 measured $0.0469 $0.0454
Channel compliance% of products, higher is better 28 · 1 of 36 measured 99.4% 100.0%
Channel rules broken% of products, lower is better 22 · 1 of 36 measured 0.6% 0.0%
DeckBench measured by Spring Prompt · 3 results
Cost per deckUS dollars, lower is better 31 · 32 of 36 measured $0.45 $0.54
Deck ratingrating, higher is betterThe rating weighs each win by how strong the other deck was; the win rate counts wins against whichever decks a model happened to meet, so the two can disagree. 16 · 14 of 36 measured 1,194 1,223
Accurate decks% of tasks, higher is better 17 · 11 of 36 measured 50.0% 66.7%
ROASBench measured by Spring Prompt · 3 results
Overall scorescore out of 100, higher is better 20 · 5 of 36 measured 44.1 55.1
Audience scorescore out of 100, higher is better 18 · 11 of 36 measured 65.1 67.6
Business scorescore out of 100, higher is better 20 · 4 of 36 measured 40.2 58.2
Artificial Analysis reported by Artificial Analysis · 3 results
Humanity's Last Exam% of questions, higher is better 20 · 9 of 250 measured 46.9%max reasoning 52.5%
Terminal-Bench 4.0% of tasks, higher is better 37 · 1 of 120 measured 12.6%low reasoning 67.2%
Long-context reasoning (AA-LCR)% of questions, higher is better 1 · 3 of 241 measured 88.7%max reasoning 86.3%
Show all 64 results (49 more)
MeasureKimi K3Kimi K3.1 Pro
BulletBench measured by Spring Prompt · 2 results
Invalid moves · Bullet 60s% of moves, lower is better 1 · 1 of 62 measured 0.0%low reasoning 0.0%low reasoning
Median move time · Bullet 60smilliseconds, lower is better 52 · 62 of 62 measured 2.6 slow reasoning 5.0 slow reasoning
CatalogBench measured by Spring Prompt · 26 results
Channel rules broken · sales brief% of products, lower is better 19 · 1 of 36 measured 0.6% 0.0%
Claims to check% of products, lower is better 1 · 1 of 34 measured 0.0% 0.0%
Claims to check · sales brief% of products, lower is better 1 · 1 of 34 measured 0.0% 0.0%
Conflicts caught% of conflicts, higher is better 34 · 1 of 36 measured 97.0% 100.0%
Content quality% of checks, higher is better 25 · 1 of 36 measured 96.4% 98.9%
Content quality · sales brief% of checks, higher is better 23 · 30 of 36 measured 96.4% 95.3%
Decision accuracy% of decisions, higher is better 23 · 1 of 36 measured 96.1% 98.9%
Failed outputs% of products, lower is better 1 · 1 of 36 measured 0.0% 0.0%
Failed outputs · sales brief% of products, lower is better 28 · 34 of 36 measured 0.6% 3.6%
Field accuracy% of missing fields, higher is better 10 · 3 of 36 measured 94.4% 95.6%
Invented values% of filled values, lower is better 35 · 26 of 36 measured 5.7% 2.1%
Missing UK information% of products, lower is better 1 · 1 of 36 measured 0.0% 0.0%
Missing UK information · sales brief% of products, lower is better 1 · 1 of 36 measured 0.0% 0.0%
Not findable% of products, lower is better 18 · 21 of 36 measured 4.2% 5.4%
Not findable · sales brief% of products, lower is better 15 · 17 of 36 measured 3.0% 3.6%
Publish-ready listings% of products, higher is better 21 · 4 of 36 measured 61.3% 78.6%
Publish-ready listings · sales brief% of products, higher is better 27 · 3 of 36 measured 14.9% 75.0%
Time per productseconds, lower is better 34 · 36 of 36 measured 66.6 114
Unsupported claimsclaims per product, lower is better 27 · 1 of 36 measured 0.98 0.00
Unsupported claims% of products, lower is better 19 · 1 of 36 measured 16.7% 0.0%
Unsupported claims · sales brief% of products, lower is better 31 · 2 of 36 measured 81.6% 1.8%
Unsupported claims · sales briefclaims per product, lower is better 30 · 7 of 36 measured 3.64 0.21
Wrong attributes% of products, lower is better 33 · 18 of 36 measured 24.4% 16.1%
Wrong attributes · sales brief% of products, lower is better 28 · 6 of 36 measured 20.8% 14.3%
Wrong category or variant% of products, lower is better 27 · 7 of 36 measured 3.0% 1.8%
Wrong category or variant · sales brief% of products, lower is better 31 · 33 of 36 measured 4.2% 5.4%
DeckBench measured by Spring Prompt · 14 results
Caveat dropped% of tasks, lower is better 1 · 16 of 36 measured 0.0% 16.7%
Clean layout% of tasks, higher is better 18 · 31 of 36 measured 33.3% 0.0%
Design quality% of the maximum, higher is better 7 · 5 of 36 measured 81.4% 83.9%
Draft figure quoted% of tasks, lower is better 1 · 1 of 36 measured 0.0% 0.0%
Findings missing% of tasks, lower is better 25 · 33 of 36 measured 16.7% 33.3%
Head-to-head win rate% of comparisons, higher is betterThe rating weighs each win by how strong the other deck was; the win rate counts wins against whichever decks a model happened to meet, so the two can disagree. 11 · 10 of 36 measured 69.8% 70.8%
Layout defects% of tasks, lower is better 18 · 31 of 36 measured 66.7% 100.0%
Misleading metric used% of tasks, lower is better 1 · 1 of 36 measured 0.0% 0.0%
Presentable decks% of tasks, higher is better 17 · 17 of 36 measured 0.0% 0.0%
Recommendation late or wrong% of tasks, lower is better 22 · 1 of 36 measured 33.3% 0.0%
Slides needing work% of tasks, lower is better 5 · 1 of 36 measured 16.7% 0.0%
Time per deckseconds, lower is better 33 · 36 of 36 measured 438 1,288
Unsupported claims% of tasks, lower is better 26 · 1 of 36 measured 33.3% 0.0%
Unsupported numbers% of tasks, lower is better 1 · 1 of 36 measured 0.0% 0.0%
ROASBench measured by Spring Prompt · 6 results
Consistency scorescore out of 100, higher is better 21 · 9 of 36 measured 28.3 36.3
Contribution profitsimulated US dollars, higher is better 16 · 4 of 36 measured $654,270 $858,023
Cost of a runUS dollars, lower is better 30 · 34 of 36 measured $1.01 $1.60
Months over budgetmonths of 12, lower is better 1 · 1 of 36 measured 0 0
Planning scorescore out of 100, higher is better 22 · 5 of 36 measured 54.7 55.0
Return on ad spendprofit per $1 spent, higher is better 21 · 2 of 36 measured $0.50 $0.72
Artificial Analysis reported by Artificial Analysis · 1 result
SciCode% of problems, higher is better 11 · 7 of 121 measured 59.5%max reasoning 60.8%

Quick answers

Is Kimi K3 better than Kimi K3.1 Pro?

Pick Kimi K3.1 Pro: it scores higher (Spring Prompt overall 67 against 46) at about the same price.

Which is cheaper, Kimi K3 or Kimi K3.1 Pro?

They cost about the same to run: $0.047 and $0.045 per product listing on CatalogBench.

Which is better for decks from an analysis, Kimi K3 or Kimi K3.1 Pro?

Kimi K3.1 Pro ranks 14 of 35 models on sale for turning a finished analysis into a deck you could present as it is, against 16 for Kimi K3.

Which is better for marketing planning, Kimi K3 or Kimi K3.1 Pro?

Kimi K3.1 Pro ranks 5 of 35 models on sale for planning a year of ad spend without overspending, against 20 for Kimi K3.

Which is better for agents and tool use, Kimi K3 or Kimi K3.1 Pro?

Kimi K3.1 Pro ranks 1 of 159 models on sale for multi-step tasks with tools: support desks, coding agents, function calls, against 40 for Kimi K3.

Which is better for reasoning and knowledge, Kimi K3 or Kimi K3.1 Pro?

Kimi K3.1 Pro ranks 9 of 187 models on sale for hard questions across science, maths and general knowledge, against 18 for Kimi K3.

Which is better for speed under pressure, Kimi K3 or Kimi K3.1 Pro?

Kimi K3.1 Pro ranks 29 of 38 models on sale for good decisions against a real clock (fast chess), against 36 for Kimi K3.

Run Kimi K3 and Kimi K3.1 Pro on your own prompt.

Benchmarks aren't your task. Test both side by side on your prompt, with every answer checked against your rules and the cost of each one.