Kimi K3.1 Pro ranks fourth on Artificial Analysis with a score of 55
The benchmark lists $2-per-million input pricing and 77.1 output tokens per second, while sources differ on the model’s proprietary and open-weight status.
By Benchdict AI DeskPublished Oct 23, 20263 min read
Kimi K3.1 Pro, an AI model accepting text and images, scores 55 against a comparable-model median of 27 on Artificial Analysis, a benchmarking site that combines 10 evaluations in its Intelligence Index. It ranks fourth among 215 models; the site lists its release month as October 2026.
Key facts
- Intelligence score
- 55 index points
- Comparable-model median
- 27 index points
- Intelligence rank
- Fourth among 215 models
- Input price
- $2 per million tokens
- Output price
- $10 per million tokens
- Output speed
- 77.1 tokens per second
The ranking and competing models
Artificial Analysis places Kimi K3.1 Pro fourth with 55 points, against a comparable-model median of 27. Its measurements cover the reasoning version—the version that works through intermediate steps before answering—and use Intelligence Index version 4.3.2.
Techmeme’s headline, summarizing an Artificial Analysis post, names the three higher-ranked rival models as Opus 5.5, Sonnet 5.5 and Muse Wonder Full, each run at its “(max)” setting.
Listed prices and measured speed
Artificial Analysis lists prices of $2 per million input tokens and $10 per million output tokens, with a 90% cache discount. Tokens are the units used to meter material sent to a model and the text it generates. The reported average Intelligence Index evaluation cost is $2.11 per task.
That task cost describes the benchmark workload rather than a fixed price for an application request. The site calculates it using multiple token-price categories and weights the result by each evaluation’s share of the index. Its evaluation generated 150 million tokens, compared with a reported median of 87 million.
Artificial Analysis measures output speed at 77.1 tokens per second, which it describes as below its reported average. The model accepts text and image inputs, produces text, and has a one-million-token context window—the capacity for material held within a request—for teams processing long documents or image-containing requests.
Conflicting labels on weights and access
Artificial Analysis labels Kimi K3.1 Pro proprietary. Techmeme calls it open-weight and says full model weights—the files needed to run the model—are due by November 4.
The benchmark page lists prices yet also displays repeated “No publicly available API” labels. An API is the interface applications use to call a hosted model.
Not yet confirmed
- No first-party announcement from the model’s developer was read. The open-weight label and November 4 weights commitment rest only on Techmeme’s summary; delivery, license terms and self-hosting requirements remain unconfirmed.
- The API labels lack enough context to identify which models or rows they describe. Provider-specific access, regional availability, rate limits and maximum output length remain unresolved.
- The three higher-ranked models and their settings are named only in Techmeme’s headline; directly comparable rival prices and speeds are unavailable.
- The chart’s “Pareto line” label does not establish whether Kimi K3.1 Pro lies on the cost-versus-intelligence frontier.
- The exact launch day, score-publication date, complete benchmark run settings and uncertainty estimates are unavailable.
- Comparable Kimi K3 baseline figures and separate score, price, speed and availability details for K3.1 Pro Highspeed are unavailable.
What to watch
- Whether full model weights arrive by November 4, with license terms and deployment requirements.
- Clarification of API access and directly comparable rival pricing and speed measurements.
Sources
- Kimi K3.1 scores 55 on Artificial Analysis, placed on the Pareto lineHacker News (front page), Oct 23, 2026.
- Artificial Analysis says Kimi K3.1 Pro lands 4th on its Intelligence Index, behind only Opus 5.5 (max), Sonnet 5.5 (max), and Muse Wonder Full (max) (@artificialanlys)Techmeme, Oct 23, 2026.
Benchdict AI Desk, AI-assisted reporting desk. Stories follow our editorial standards; corrections are logged above and on the corrections page.