Skip to content
aiqbench

Kimi · model passport

MoonshotAI: Kimi K2 Thinking

kimi-k2-thinking
  • Open weights
  • 1.03T params
  • 262,144 context
  • Released 2025-11-04(HF repo created)
5/27axes covered
Best percentile
95.8LiveCodeBench (Artificial Analysis)
Cohort standing
66.3median 53.6

What the data says

Benchmarked on 5 of 27 tracked axes. Every figure below is the model's own published result on that board.

Strongest axes

  • 95.8 pct — LiveCodeBench (Artificial Analysis) (85.3%)
  • 89.9 pct — MMLU-Pro (Artificial Analysis) (84.8%)

Uneven: 54.9 percentile points between its best and worst axis — 95.8 pct on LiveCodeBench (Artificial Analysis) versus 40.9 pct on Terminal-Bench.

  • LiveCodeBench (Artificial Analysis)95.8 pct85.3%Verify →
  • MMLU-Pro (Artificial Analysis)89.9 pct84.8%Verify →
  • IFBench (Artificial Analysis)79 pct68.1%Verify →
  • LiveBench74.6 pct65.8%Verify →
  • Terminal-Bench40.9 pct35.7%Verify →
  • OverallNot benchmarked
  • WebDev ArenaNot benchmarked
  • Vision ArenaNot benchmarked
  • Factuality ArenaNot benchmarked
  • Search ArenaNot benchmarked
  • Document ArenaNot benchmarked
  • Agent ArenaNot benchmarked
  • GPQA DiamondNot benchmarked
  • AIME (math)Not benchmarked
  • FrontierMathNot benchmarked
  • SimpleQA VerifiedNot benchmarked
  • ARC-AGI-2Not benchmarked
  • Humanity's Last ExamNot benchmarked
  • SimpleBenchNot benchmarked
  • BFCL v4 (function calling)Not benchmarked
  • SWE-bench Verified (live board)Not benchmarked
  • Terminal-Bench 4.0Not benchmarked
  • HHEM Hallucination (Vectara)Not benchmarked
  • SWE-bench VerifiedNot benchmarked
  • CodingNot benchmarked
  • Instruction followingNot benchmarked
  • Hard promptsNot benchmarked

Percentiles are the model's standing within each board's own population. Axes marked "not benchmarked" have no published result for this model — that is an absence of data, not a low score.

How to run it

Run it yourself

Repositorymoonshotai/Kimi-K2-Thinking
Parameters1.03T
ArchitectureDeepseekV3ForCausalLM

Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "moonshotai/Kimi-K2-Thinking"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

Ollama

We don't sync Ollama tags, so we won't print a command we can't verify — a wrong one would pull the wrong model. Search the Ollama library →

Unsloth fine-tuning recipes are model-specific and we don't derive them here. Unsloth docs →

Use it hosted

API model IDmoonshotai/kimi-k2-thinking
Context length262,144 tokens
Max output98,304 tokens
Input price$0.60 per 1M tokens
Cached input price$0.15 per 1M tokens
Output price$2.50 per 1M tokens

API parameters this endpoint supports

Frequency penaltyInclude reasoningLogprobsMax tokensPresence penaltyReasoningRepetition penaltyResponse formatSeedStop sequencesStructured outputsTemperatureTool choiceTool callingTop-kTop logprobsTop-p

Example request

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"moonshotai/kimi-k2-thinking","messages":[{"role":"user","content":"Hello"}]}'

View on OpenRouter →

Identifiers, limits and prices are synced from the upstream source shown above. Any field we haven't synced is left out entirely rather than guessed.

  • kimi-k2-thinking stands at 66.3 (raw mean 76) — measured on 5 of 27 axes.

    standingtemplateVerify →

  • kimi-k2-thinking stands at 66.3 versus a cohort median of 53.6 across 132 models released within 180 days — above the peer median.

    cohorttemplateVerify →

Every claim is a computed fact from the current board data, phrased by a low-temperature model call and rejected by a validation gate if any number is not in the source row. How the leaderboards work →