Kimi K3.1 Pro — Benchmarks

Benchmark scores for Kimi K3.1 Pro aggregated from public leaderboards (4 boards), with how it ranks among open models and how it compares with its nearest neighbours. See hardware requirements for what you need to run it.

llmrun Composite Score →

142.8

90% range 137.5–148.2 · from 4 boards

Coding140.790% range 131–150
AgentsNot enough data yet
MathNot enough data yet
Science138.090% range 126–150
ReasoningNot enough data yet

Kimi K3.1 Pro reasoning benchmarks

Multi-step logic and problem-solving — can the model think through a hard problem rather than recall an answer.

CritPt
CritPt (Critical Points) tests graduate-level physics problem-solving on research-style questions spanning mechanics, electromagnetism, quantum mechanics, and more. It was created by physicists to probe whether a model can reason through multi-step physics rather than pattern-match textbook answers.
24.6%
24.6%
median 0.9%open leader 26.6%
Overall best: Gemini 4 Carbon (high) 36.0%
99th percentile · #2 of 84 open · #51 of 244 all
#2 of 84
#51 of 244
Lab-reported
Epoch AI · Oct 2026

Kimi K3.1 Pro coding benchmarks

Writing, editing, and fixing real code — measured by whether the resulting program actually runs and passes tests.

SciCode
SciCode asks a model to write real scientific-computing code — implementing multi-step numerical methods drawn from physics, chemistry, biology, and math research — and checks the output against reference solutions. It was built by a team of scientists and ML researchers to test coding ability on genuine research problems, not typical software tasks.
60.8%
60.8%
median 40.1%open leader 60.9%
Overall best: Claude Opus 5.5 Max 66.9%
99th percentile · #2 of 72 open · #10 of 217 all
#2 of 72
#10 of 217
Lab-reported
Epoch AI · Oct 2026
ALE-Bench
ALE-Bench, from Sakana AI and AtCoder, tests long-horizon algorithmic optimisation using problems drawn from AtCoder Heuristic Contests, where there is no single correct answer and solutions are scored on quality. Results are reported as an AtCoder-style performance rating, where higher is better.
1932.8
1932.8
median 754.6open leader
Overall best: GPT 6 Astra Max 2951.3
top · #1 of 48 open · #13 of 129 all
#1 of 48
#13 of 129
Lab-reported
Epoch AI · Oct 2026

Kimi K3.1 Pro math benchmarks

Symbolic and quantitative problem-solving, from competition math to multi-step calculation.

ProofBench
ProofBench evaluates a model's ability to write and check rigorous mathematical proofs, rather than just produce a final numeric answer. It probes step-by-step logical correctness, which is a harder and more distinct skill than solving for a number.
91.0%
91.0%
median 18.0%open leader
Overall best: Claude Fable 5.1 Max 100.0%
top · #1 of 31 open · #9 of 84 all
#1 of 31
#9 of 84
Lab-reported
Epoch AI · Oct 2026

Score vs VRAM in the ≤2 TB tier

Kimi K3.1 Pro
Other models
Best Score for its VRAM
x: VRAM (GB, log scale) · y: llmrun Score

#1 of 106 that fit ≤2 TB

Each dot is a model at the best quant that fits ≤2 TB. The line steps up to the best Score available at each amount of VRAM; a dot below it is beaten by a model that needs no more memory.

Full ≤2 TB ranking →

Kimi K3.1 Pro vs its neighbours

Same boards for the nearest open models per llmrun Score. The highest score in each row is marked.

Board	Kimi K3.1 Prothis model	Kimi K3nearest by ScoreQ4_K_M · 1670 GB	DeepSeek V4 Pro 0813nearest by ScoreQ4_K_M · 991 GB	DeepSeek V4 Pronearest by ScoreQ4_K_M · 960 GB
llmrun Score	142.8	132.7	129.2	116.7
CritPt	24.6%	23.4%	18.0%	12.9%
SciCode	60.8%	59.5%	51.0%	50.0%
ALE-Bench	1932.8	1524.5	1403.2	1006.1
ProofBench	91.0%	87.0%	50.0%	16.0%
Compare head to head		Compare	Compare	Compare

Scores aggregated from public benchmark sources (Epoch AI: 4 boards) — each linked from the benchmark pages. llmrun does not run these benchmarks. Percentiles and ranks are among open models in the llmrun catalogue; the median and leader marks refer to the same field.