Kimi K3.1 ranks well for intelligence and sits on the cost-efficiency line

What you could do with it

• A legal professional could feed the model up to 1,500 A4 pages of documents at once and receive detailed text analysis to support case research ¹.
• An accountant could use the model to work through complex finance and accounting tasks, relying on its strong reasoning ability despite longer response times ¹.
• An engineering team could deploy the model for challenging coding, automation, and scientific problem-solving work where deep thinking matters more than speed ¹.

What the scores say

Kimi K3.1 Pro ranked #4 out of 215 models on the Artificial Analysis Intelligence Index with a score of 55, far above the median score of 27 ¹. The model was tested by generating 113 million tokens of output, making it somewhat verbose compared to the median of 104 million .

In terms of speed, the model produced 77.1 output tokens per second, placing it at #106 out of 186 models — slower than the average of 81 tokens per second ¹.

On cost, the model charges $2.00 per million input tokens (expensive, median: $0.43) and $10.00 per million output tokens (expensive, median: $1.84) ¹. It also offers a 90 percent cache discount . The average cost per Intelligence Index task came out to $3.68 .

Being on the Pareto line

Artificial Analysis plots every model on a chart that shows intelligence against the cost per task. Models on the "Pareto line" are the ones where you cannot get more intelligence without paying more ¹.

Kimi K3.1 Pro sits on this efficiency frontier alongside models from companies like Google, OpenAI, Anthropic, DeepSeek, and StepFun ¹. This means the model is considered one of the best value options for its level of intelligence — you are not paying extra for smarts that could be had cheaper elsewhere .

What the model can do

Kimi K3.1 Pro is an open weights reasoning model released in October 2026 ¹. It supports text and image input and outputs text only . It has a context window of one million tokens, which is roughly equivalent to about 1,500 A4 pages of size 12 Arial font .

The Intelligence Index itself is made up of 9 different evaluations: AA-Briefcase v1.1, AutomationBench-AA, Terminal-Bench 4.0, FrontierCode-AA 2.0, AA-Omniscience v1.1, GDP.pdf, HLE-Diamond, CritPt v2 (Hard Subset), and Terminal-Bench-Science 0.1 ¹. These tests measure skills in areas like coding, automation, science, finance, legal work, and general problem solving .

While Kimi K3.1 Pro delivers top-tier intelligence for its price range, users should expect longer responses and slower output speeds compared to many other models on the market ¹.

Who is it for?

The model is aimed at users who need strong reasoning ability and can work with longer generation times and wordier outputs ¹. It is particularly suited for tasks in finance, accounting, legal work, engineering, and strategy — areas where deep thinking matters more than speed .

The model currently has no publicly available API, which means access may be limited to certain platforms or partnerships at this time ¹.

Corrections and updates

• Corrected on 12 November 2026. An earlier version described Kimi K3.1 Pro as proprietary, at median pricing, slower-performing, and more verbose; the corrected version states it is open weights, well above median on both input and output prices, moderately verbose, and costly.

Key facts

• Kimi K3.1 Pro scored 55 on the Artificial Analysis Intelligence Index, ranking #4 out of 215 models
• It sits on the Pareto efficiency line, meaning it delivers strong intelligence for its price
• The model costs $2 per million input tokens and ## Key facts

0 per million output tokens, both well above the median (input median: $0.43, output median: ## Key facts

.84)

• It generates 113 million tokens during testing, making it somewhat verbose compared to the 104M median
• Speed is 77.1 tokens per second, slower than the average of 81

Sources

1. Kimi K3.1 Pro - Intelligence, Performance & Price Analysis | Artificial Analysis