Kimi K2 Instruct · Aider coding benchmarks · Polyglot · percent correct ↗
Kimi / MODEL EVIDENCE
Kimi K2 Instruct
22 benchmarks · 25 observations · 23 sources · released 2025-07-12
Scores at a glance
22 benchmarks · large dot: Kimi K2 Instruct; small dots: closest models where reported
Fiction.liveBench ↗61.1%
Kimi K2 Instruct · Fiction.LiveBench · 16k token score · % ↗
GSO ↗4.9%
Kimi K2 Instruct · GSO-Bench · Score OPT@1 · % ↗
Lech Mazur Writing ↗8.56 / 10
Kimi K2 Instruct · Lech Mazur Writing · Mean rubric score / 10 ↗
SimpleBench ↗26.3%
Kimi K2 Instruct · SimpleBench · Score (AVG@5) · % ↗
Terminal-Bench 2.0 ↗27.79%
Kimi K2 Instruct · Terminal Bench · Accuracy mean · % ↗
3 more reported results
26.74% · Kimi K2 Instruct · Terminal Bench · Accuracy mean · % · Epoch AI · Terminal Bench ↗
27.8% · Kimi K2 Instruct · Terminal Bench · Accuracy mean · % · Epoch AI · Terminal Bench ↗
26.7% · Kimi K2 Instruct · Terminal Bench · Accuracy mean · % · Epoch AI · Terminal Bench ↗
WeirdML (v2) ↗39.36%
Kimi K2 Instruct · WeirdML · Accuracy · % ↗
LegalBench ↗81.45 %
Kimi K2 Instruct · LegalBench · Accuracy ↗
CorpFin v2 ↗50.39 %
Kimi K2 Instruct · CorpFin · Accuracy ↗
Kimi K2 Instruct · TaxEval · Accuracy ↗
MedQA (USMLE) ↗83.97 %
Kimi K2 Instruct · MedQA · Accuracy ↗
AIME 2024–2025 · Vals ↗62.71 %
Kimi K2 Instruct · AIME · Accuracy ↗
MATH-500 ↗94.2 %
Kimi K2 Instruct · MATH 500 · Accuracy ↗
MGSM (Multilingual GSM) ↗90.95 %
Kimi K2 Instruct · MGSM · Accuracy ↗
GPQA Diamond ↗71.46 %
Kimi K2 Instruct · GPQA Diamond · Accuracy ↗
MMLU-Pro ↗79.39 %
Kimi K2 Instruct · MMLU Pro · Accuracy ↗
LiveCodeBench ↗70.45 %
Kimi K2 Instruct · LiveCodeBench · Accuracy ↗
kimi-k2-instruct · Scale Labs · Score ↗
kimi-k2-instruct · Scale Labs · Score ↗
Fortress ↗55.47%
kimi-k2-instruct · Scale Labs · Risk elicitation score ↗
kimi-k2-instruct · Scale Labs · Score ↗
ForecastBench ↗62.2 Brier index
kimi-k2-instruct† · Overall difficulty-adjusted Brier index ↗
Closest models by score
Similar positions on shared benchmark leaderboards. Smaller gaps are closer.
Qwen 3 (235B) →Qwen
9 of 14 shared benchmark comparisons
9 of 14 shared benchmark comparisons
DeepSeek V3 (03/24/2025) →DeepSeek
10 of 14 shared benchmark comparisons
How similarity is measured
Mean absolute gap in percentile positions on shared benchmark settings; lower is closer. At least 60% overlap is required, with a minimum of three when available. Independent evaluations are used when there are at least three; otherwise published report comparisons are used. Repeated model observations within one setting are averaged before ranking. Writing-style traits and composite indices are excluded. This describes similarity, not an overall quality ranking.
Sources
Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.
- Aider LLM Leaderboards — Polyglot coding benchmark ↗
- Forecastbench — methodology and published results ↗
- SimpleBench ↗
- Terminal-Bench 2.0 ↗
- Fiction.liveBench — March 14, 2025 benchmark release ↗
- Writing benchmark — incorporating mandatory story elements ↗
- GSO: Challenging Software Optimization Tasks for Evaluating SWE-Agents ↗
- WeirdML ↗
- Scale Labs — evaluation methodology and results ↗
- Scale Labs — evaluation methodology and results ↗
- Scale Labs — evaluation methodology and results ↗
- Scale Labs — evaluation methodology and results ↗
- Terminal-Bench 2.0 leaderboard ↗
- AIME — evaluation methodology and results ↗
- CorpFin — evaluation methodology and results ↗
- GPQA Diamond — evaluation methodology and results ↗
- LiveCodeBench — evaluation methodology and results ↗
- LegalBench — evaluation methodology and results ↗
- MATH 500 — evaluation methodology and results ↗
- MedQA — evaluation methodology and results ↗
- MGSM — evaluation methodology and results ↗
- MMLU Pro — evaluation methodology and results ↗
- TaxEval — evaluation methodology and results ↗
Names used in source tables
Kimi K2 Instruct · kimi-k2-instruct · kimi-k2-instruct†