Kimi K3.1 Pro: score and benchmark evidence
Historical release · · Moonshot AI
136.1 points · Rank 6 among 38 ranked models · Stability range 118.5–149.6 points.
Rank follows the point estimate; it does not establish statistically significant superiority. The scale is anchored at mean 100, SD 15 in the frozen calibration cohort, not human IQ or a percentage.
One score summarizes demonstrated capability across published reasoning settings and seven equally weighted domains. Related benchmark variants share a family budget, and influence from one evaluator is capped.
Release tli-2026-v1.0-2026-10-28-refresh-44. Methodology and limitations for this release.
This page preserves historical evidence. See the current model record.
Capability domains
Domain scores use calibration-cohort standard-deviation units. Unmeasured domains are shown as unavailable, not zero.
| Domain | Score (SD units) | Observed benchmarks |
|---|---|---|
| Knowledge & reasoning | 0.96 | 5 |
| Coding & software engineering | 1.04 | 4 |
| Agentic tool & computer use | 1.19 | 3 |
| Professional & real-world work | 0.98 | 4 |
| Multimodal & vision | 0.98 | 2 |
| Cybersecurity | Unavailable | 0 |
| Preference & communication | Unavailable | 0 |
Published benchmark evidence
18 results contribute to this score, from 19 collected results and 17 contributing benchmark families. Counts are not independent sample sizes.
Results, source citations, review decisions and compatible reasoning modes are stored in a versioned relational database. Each publication uses a sealed, checksum-verified export; CSV files are optional exports, not editable scoring inputs. Previous revisions remain available for audit. Collected means a numeric result exists in the database. Used means its source, exact model checkpoint and benchmark protocol have been reviewed and its benchmark has a frozen calibration supported by multiple providers. For each benchmark we select the highest published aggregate across comparable reasoning modes, not necessarily maximum effort, and count it once. More tested settings provide more chances to record a high score; this selection advantage is not corrected by the conditional interval. Developer-reported and sole/unspecified settings remain eligible and disclosed. Unresolved does not mean false or useless: conflicting metrics, unverified citations and unmatched versions stay collected until reconciled.
The score stability range is a 95% bootstrap score interval conditional on frozen calibration and comparable evidence availability. It resamples benchmark families and evaluators and adds a frozen missing-evidence correction. It excludes selection across reasoning settings, selective publication, calibration changes and new domains. It is not a 95% guarantee about true capability or where a future score will land. Sparse or conflicting evidence can widen the range.
Reporting sensitivity: ±14.2 points. Sensitivity to selectively published results. This stress range is not a confidence interval or a probability statement.
AA-Briefcase v1.1 (Elo): 1677.79 native-points
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (none); Artificial Analysis independent evaluation, 2026-10-26; fallback not reported
Metric: published native-points
Original source · Retrieved 2026-10-26
Exact model and effort in the evaluator chart payload.
GDPval-AA v2.1 (Elo): 1845.72 native-points
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (none); Artificial Analysis independent evaluation, 2026-10-26; fallback not reported
Metric: published native-points
Original source · Retrieved 2026-10-26
Exact model and effort in the evaluator chart payload.
AutomationBench-AA: 73.2977109092%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (none); Artificial Analysis independent evaluation, 2026-10-26; fallback not reported
Metric: published percent
Original source · Retrieved 2026-10-26
Exact model and effort in the evaluator chart payload.
Terminal-Bench 4.0 — AA harness: 67.1717171717%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (none); Artificial Analysis independent evaluation, 2026-10-26; fallback not reported
Metric: published percent
Original source · Retrieved 2026-10-26
Exact model and effort in the evaluator chart payload.
SciCode: 60.7638888889%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (none); Artificial Analysis independent evaluation, 2026-10-26; fallback not reported
Metric: published percent
Original source · Retrieved 2026-10-26
Exact model and effort in the evaluator chart payload.
HLE text-only — AA harness: 52.4559777572%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (none); Artificial Analysis independent evaluation, 2026-10-26; fallback not reported
Metric: published percent
Original source · Retrieved 2026-10-26
Exact model and effort in the evaluator chart payload.
GDP.pdf — AA harness: 31.4%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (none); Artificial Analysis independent evaluation, 2026-10-26; fallback not reported
Metric: published percent
Original source · Retrieved 2026-10-26
Exact model and effort in the evaluator chart payload.
CritPt: 24.5714285714%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (none); Artificial Analysis independent evaluation, 2026-10-26; fallback not reported
Metric: published percent
Original source · Retrieved 2026-10-26
Exact model and effort in the evaluator chart payload.
AA-Omniscience: 29.866667 native-points
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (none); Artificial Analysis independent evaluation, 2026-10-28; fallback not reported
Metric: published native-points
Original source · Retrieved 2026-10-28
Daily evaluator refresh; replaces 29.5166666667 recorded 2026-10-26.
AA-LCR v1.1: 86.3333333333%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (none); Artificial Analysis independent evaluation, 2026-10-26; fallback not reported
Metric: published percent
Original source · Retrieved 2026-10-26
Exact model and effort in the evaluator chart payload.
MMMU-Pro: 83.8728323699%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (none); Artificial Analysis independent evaluation, 2026-10-26; fallback not reported
Metric: published percent
Original source · Retrieved 2026-10-26
Independent MMMU-Pro aggregate for this effort.
AA Intelligence Index v4.3.2 (score): 46.002624 native-points
Not used in the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (none); Artificial Analysis independent evaluation, 2026-10-28; fallback not reported
Metric: published native-points
Original source · Retrieved 2026-10-28
Daily evaluator refresh; replaces 55.0869210959 recorded 2026-10-26.
Terminal-Bench 4.0: 65.7%
Contributes to the score. Developer-reported. Review status: reviewed-developer.
Configuration: Kimi K3.1 Pro (max); Kimi K3.1 Pro developer-reported result (moonshotai); fallback not reported
Metric: published percent
Original source · Retrieved 2026-10-26
Developer-reported result entered in the newsroom. Same protocol as the catalogue row: the lab table reproduces cells we hold (gpt-6-astra 58.2, opus-5-5 66.4). Claude-Code harness (Footnotes, Terminal-Bench 4.0).
BrowseComp: 92.1%
Contributes to the score. Developer-reported. Review status: reviewed-developer.
Configuration: Kimi K3.1 Pro (max); Kimi K3.1 Pro developer-reported result (moonshotai); fallback not reported
Metric: published percent
Original source · Retrieved 2026-10-26
Developer-reported result entered in the newsroom. Same protocol as the catalogue row: the lab table reproduces cells we hold (gpt-6-astra 91.5). Context compaction at 200K tokens, max 500 steps, temperature 1.0, top-p 1.0 (Footnotes, BrowseComp).
BabyVision (with CI): 95.4%
Contributes to the score. Developer-reported. Review status: reviewed-developer.
Configuration: Kimi K3.1 Pro (max); Kimi K3.1 Pro developer-reported result (moonshotai); fallback not reported
Metric: published percent
Original source · Retrieved 2026-10-26
Developer-reported result entered in the newsroom. Same protocol as the catalogue row: the lab table reproduces cells we hold (deepseek-v4-1-flash-1017 92.5). Lab labels the row BabyVision w/ python; averaged over four runs (Footnotes, multimodal benchmarks w/ python).
BioMysteryBench v1 — Vals: 75.56%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (max); Vals AI BioMysteryBench harness; provider Moonshot AI; fallback not reported
Metric: published percent
Original source · Retrieved 2026-10-27
Vals AI model page, headline accuracy 75.56 ± 0.64, rank 7/29. Vals notes some benchmarks may use a different provider or parameters.
Legal Research Bench — Vals: 37.5%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (max); Vals AI Legal Research Bench harness; provider Moonshot AI; fallback not reported
Metric: published percent
Original source · Retrieved 2026-10-27
Vals AI model page, headline accuracy 37.5 ± 3.37, rank 42/86. Vals notes some benchmarks may use a different provider or parameters.
ProofBench v1.1 — Vals: 91%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (max); Vals AI ProofBench v1.1 harness; provider Moonshot AI; fallback not reported
Metric: published percent
Original source · Retrieved 2026-10-27
Vals AI model page, headline accuracy 91.0 ± 2.88, rank 10/52. Vals notes some benchmarks may use a different provider or parameters.
Vibe Code Bench v1.1 — Vals: 84.89%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: Kimi K3.1 Pro (max); Vals AI Vibe Code Bench v1.1 harness; provider Moonshot AI; fallback not reported
Metric: published percent
Original source · Retrieved 2026-10-27
Vals AI model page, headline accuracy 84.89 ± 2.95, rank 17/120. Vals notes some benchmarks may use a different provider or parameters.
Cite this model record
The Latent. “Kimi K3.1 Pro: score and benchmark evidence.” 2026-10-28. Release tli-2026-v1.0-2026-10-28-refresh-44.
Download release data and definitions (JSON) · Data usage terms