Kimi K2.6 (thinking) · HLE-Full(w/ tools) ↗
1 more reported result
34.7 · Kimi K2.6 (thinking) · HLE-Full · Kimi K2.6 technical model card ↗
Kimi / MODEL EVIDENCE
Kimi K2.6
69 benchmarks · 87 observations · 54 sources · released 2026-04-20
Scores at a glance
69 benchmarks · large dot: Kimi K2.6; small dots: closest models where reported
BrowseComp ↗83.2
Kimi K2.6 (thinking) · BrowseComp ↗
DeepSearchQA ↗92.5
Kimi K2.6 (thinking) · DeepSearchQA(f1-score) ↗
1 more reported result
83.0 · Kimi K2.6 (thinking) · DeepSearchQA(accuracy) · Kimi K2.6 technical model card ↗
WideSearch ↗80.8
Kimi K2.6 (thinking) · WideSearch (item-f1) ↗
Toolathlon ↗50.0
Kimi K2.6 (thinking) · Toolathlon ↗
MCPMark ↗55.9
Kimi K2.6 (thinking) · MCPMark ↗
Claw Eval ↗62.3
Kimi K2.6 (thinking) · Claw Eval (pass^3) ↗
1 more reported result
80.9 · Kimi K2.6 (thinking) · Claw Eval (pass@3) · Kimi K2.6 technical model card ↗
Kimi K2.6 (thinking) · APEX-Agents ↗
Kimi K2.6 (thinking) · OSWorld-Verified ↗
Kimi K2.6 (thinking) · Terminal-Bench 2.0(Terminus-2) ↗
SWE-bench Pro ↗58.6
Kimi K2.6 (thinking) · SWE-Bench Pro ↗
Kimi K2.6 (thinking) · SWE-Bench Multilingual ↗
SWE-bench Verified ↗76.65%
Kimi K2.6 · SWE-Bench verified · Best score (across scorers) · % ↗
2 more reported results
76.2 % · Kimi K2.6 (Mini-SWE-agent) · SWE-bench Verified · Score · Vals AI · SWE-bench Verified ↗
80.2 · Kimi K2.6 (thinking) · SWE-Bench Verified · Kimi K2.6 technical model card ↗
SciCode ↗53.47%
Kimi K2.6 · SciCode · Score · % ↗
1 more reported result
52.2 · Kimi K2.6 (thinking) · SciCode · Kimi K2.6 technical model card ↗
OJBench ↗60.6
Kimi K2.6 (thinking) · OJBench (python) ↗
LiveCodeBench ↗86.77 %
Kimi K2.6 · LiveCodeBench · Accuracy ↗
1 more reported result
89.6 · Kimi K2.6 (thinking) · LiveCodeBench (v6) · Kimi K2.6 technical model card ↗
AIME 2026 ↗96.4
Kimi K2.6 (thinking) · AIME 2026 ↗
Kimi K2.6 (thinking) · HMMT 2026 (Feb) ↗
Kimi K2.6 (thinking) · IMO-AnswerBench ↗
GPQA Diamond ↗90.78%
Kimi K2.6 · GPQA diamond · Best score (across scorers) · % ↗
2 more reported results
89.14 % · Kimi K2.6 · GPQA Diamond · Accuracy · Vals AI · GPQA Diamond ↗
90.5 · Kimi K2.6 (thinking) · GPQA-Diamond · Kimi K2.6 technical model card ↗
MMMU-Pro ↗86.3 %
Kimi K2.6 · MMMU Pro · Accuracy ↗
2 more reported results
79.4 · Kimi K2.6 (thinking) · MMMU-Pro · Kimi K2.6 technical model card ↗
80.1 · Kimi K2.6 (thinking) · MMMU-Pro (w/ python) · Kimi K2.6 technical model card ↗
CharXiv ↗80.4
Kimi K2.6 (thinking) · CharXiv (RQ) ↗
1 more reported result
86.7 · Kimi K2.6 (thinking) · CharXiv (RQ) (w/ python) · Kimi K2.6 technical model card ↗
MathVision ↗87.4
Kimi K2.6 (thinking) · MathVision ↗
1 more reported result
93.2 · Kimi K2.6 (thinking) · MathVision (w/ python) · Kimi K2.6 technical model card ↗
BabyVision ↗39.8
Kimi K2.6 (thinking) · BabyVision ↗
1 more reported result
68.5 · Kimi K2.6 (thinking) · BabyVision (w/ python) · Kimi K2.6 technical model card ↗
V* ↗96.9
Kimi K2.6 (thinking) · V* (w/ python) ↗
Kimi K2.6 · FrontierMath-2025-02-28-Private · Best score (across scorers) · % ↗
Kimi K2.6 · FrontierMath-Tier-4-2025-07-01-Private · Best score (across scorers) · % ↗
Kimi K2.6 · FrontierMath-Tiers-1-3-v2-Private · Best score (across scorers) · % ↗
Kimi K2.6 · FrontierMath-Tier-4-v2-Private · Best score (across scorers) · % ↗
Kimi K2.6 · OTIS Mock AIME 2024-2025 · Best score (across scorers) · % ↗
SimpleQA Verified ↗34.9%
Kimi K2.6 · SimpleQA Verified · Best score (across scorers) · % ↗
Kimi K2.6 · Chess Puzzles · Best score (across scorers) · % ↗
Kimi K2.6 · EBR-bench · Best score (across scorers) · % ↗
Kimi K2.6 · Mystery Game Puzzles · Best score (across scorers) · % ↗
1 more reported result
12% · Kimi K2.6 · Mystery Game Puzzles · Best score (across scorers) · % · Epoch AI · Mystery Game Puzzles ↗
WeirdML (v2) ↗55.86%
Kimi K2.6 · WeirdML · Accuracy · % ↗
GBAEval ↗0.86%
Kimi K2.6 · GBAEval · Overall score · % ↗
CritPt ↗8%
Kimi K2.6 · CritPt · Accuracy · % ↗
ExploitBench ↗18.4%
Kimi K2.6 · ExploitBench · Mean capability · % ↗
1 more reported result
16.3% · Kimi K2.6 · ExploitBench · Mean capability · % · Epoch AI · ExploitBench ↗
OSWorld 2.0 ↗4.6%
Kimi K2.6 · OSWorld 2.0 · Binary accuracy · % ↗
ProofBench ↗16%
Kimi K2.6 · ProofBench · Accuracy · % ↗
LMCA ↗37.26%
Kimi K2.6 · LMCA · Score · % ↗
DTBench ↗90.93%
Kimi K2.6 · DTBench · Accuracy · % ↗
Kimi K2.6 · Furniture Assembly · Best score (across scorers) · % ↗
GDP.pdf ↗12%
Kimi K2.6 · GDP.pdf · GDP.pdf score · % ↗
Text Arena (Coding) ↗1508.77 Elo
Kimi K2.6 · WebDev Arena · Elo ↗
Kimi K2.6 · Blueprint-Bench 2 · Score · % ↗
Epoch Capabilities Index (ECI) ↗150.94 ECI
Kimi K2.6 · General ECI · 26 October 2026 fit ↗
Vals Multimodal Index ↗56.43 %
Kimi K2.6 · Vals Multimodal Index · Accuracy ↗
Legal Research Bench ↗15.87 %
Kimi K2.6 · Legal Research Bench · All-Pass ↗
LegalBench ↗84.74 %
Kimi K2.6 · LegalBench · Accuracy ↗
Kimi K2.6 · Harvey's Legal Agent Benchmark v1 · Accuracy ↗
CaseLaw v2 ↗61.2 %
Kimi K2.6 · CaseLaw · Accuracy ↗
Tax Agent Bench ↗12.4 %
Kimi K2.6 · Tax Agent Bench · Accuracy ↗
Vals Finance Agent v2 ↗44.9 %
Kimi K2.6 · Finance Agent v2 · Accuracy ↗
Kimi K2.6 · Excel Modeling Benchmark v1 · Accuracy ↗
CorpFin v2 ↗66.74 %
Kimi K2.6 · CorpFin · Accuracy ↗
MortgageTax & TaxEval v2 ↗65.82 %
Kimi K2.6 · MortgageTax · Accuracy ↗
1 more reported result
74.65 % · Kimi K2.6 · TaxEval · Accuracy · Vals AI · TaxEval ↗
MedCode ↗40.14 %
Kimi K2.6 · MedCode · Accuracy ↗
MedScribe ↗78.15 %
Kimi K2.6 · MedScribe · Accuracy ↗
MMLU-Pro ↗87.57 %
Kimi K2.6 · MMLU Pro · Accuracy ↗
Kimi K2.6 · SAGE · Accuracy ↗
Vibe Code Bench ↗37.89 %
Kimi K2.6 (OpenHands) · Vibe Code Bench · original task set ↗
Kimi K2.6 · ProgramBench · Accuracy ↗
Code Migration v1.1 ↗27.77 %
Kimi K2.6 · Code Migration v1 · Accuracy ↗
Public Benefits Bench v1.1 ↗56.63 %
Kimi K2.6 · Public Benefits Bench · Accuracy ↗
1 more reported result
53.38 % · Kimi K2.6 · Public Benefits Bench v1 · Accuracy · Vals AI · Public Benefits Bench v1 ↗
DistressBench ↗75.18%
kimi-k2.6 · DistressBench · Score ↗
Kimi k2.6 · HiL-Bench (Human-in-Loop Benchmark) · Score ↗
ALE-Bench ↗1092.67 Performance points
kimi-k2.6 · Average contest performance ↗
Vending-Bench 2 ↗6204.57 USD
Kimi K2.6 · End-of-year balance ↗
Closest models by score
Similar positions on shared benchmark leaderboards. Smaller gaps are closer.
Claude Opus 4.6 →Anthropic
22 of 33 shared benchmark comparisons
Gemini 3.5 Flash →Google
29 of 33 shared benchmark comparisons
Gemini 3.6 Flash →Google
25 of 33 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.
Reports featuring this model
Sources
Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.
- Blueprint-Bench 2 | Andon Labs ↗
- Arena leaderboard — WebDev evaluation ↗
- SciCode Benchmark | Artificial Analysis ↗
- DTBench — Conceptual Reasoning Index ↗
- LMCA — Conceptual Reasoning Index ↗
- Ale Bench — methodology and published results ↗
- Chess Puzzles ↗
- CritPt v1 ↗
- Earthborne Rangers (EBR-bench) ↗
- ExploitBench v0.1 ↗
- FrontierMath Tier 4 (v1, old) ↗
- FrontierMath Tier 4 (v2) ↗
- FrontierMath Tiers 1-3 (v1, old) ↗
- FrontierMath Tiers 1-3 (v2) ↗
- Furniture Assembly Benchmark ↗
- GPQA Diamond ↗
- Mystery Game Puzzles ↗
- OTIS Mock AIME 2024-2025 ↗
- SimpleQA Verified ↗
- SWE-bench Verified ↗
- Vending Bench 2 — methodology and published results ↗
- AI benchmarks & capabilities ↗
- Kimi K2.6 — GBA Eval ↗
- WeirdML ↗
- moonshotai/Kimi-K2.6 · Hugging Face ↗
- DistressBench — evaluation methodology and results ↗
- HiL-Bench (Human-in-Loop Benchmark) — evaluation methodology and results ↗
- OSWorld 2.0: Benchmarking computer-use agents on long-horizon real-world tasks ↗
- GDP.pdf Benchmark | Surge AI ↗
- CaseLaw — evaluation methodology and results ↗
- Code Migration v1 — evaluation methodology and results ↗
- CorpFin — evaluation methodology and results ↗
- Excel Modeling Benchmark v1 — evaluation methodology and results ↗
- Finance Agent v2 — evaluation methodology and results ↗
- GPQA Diamond — evaluation methodology and results ↗
- Harvey's Legal Agent Benchmark v1 — evaluation methodology and results ↗
- LiveCodeBench — evaluation methodology and results ↗
- LegalBench — evaluation methodology and results ↗
- Legal Research Bench — evaluation methodology and results ↗
- MedCode — evaluation methodology and results ↗
- MedScribe — evaluation methodology and results ↗
- MMLU Pro — evaluation methodology and results ↗
- MMMU Pro — evaluation methodology and results ↗
- MortgageTax — evaluation methodology and results ↗
- ProgramBench — evaluation methodology and results ↗
- ProofBench — evaluation methodology and results ↗
- Public Benefits Bench — evaluation methodology and results ↗
- Public Benefits Bench v1 — evaluation methodology and results ↗
- SAGE — evaluation methodology and results ↗
- SWE-bench Verified — evaluation methodology and results ↗
- Tax Agent Bench — evaluation methodology and results ↗
- TaxEval — evaluation methodology and results ↗
- Vals Multimodal Index — evaluation methodology and results ↗
- Vibe Code Bench — evaluation methodology and results ↗
Names used in source tables
Kimi K2.6 · Kimi K2.6 (Mini-SWE-agent) · Kimi K2.6 (OpenHands) · Kimi K2.6 (thinking) · Kimi k2.6 · kimi-k2.6