modelbench日本語

Kimi K3

Moonshot AICurrent

Input price
—
Per 1 million tokens (USD)
Output price
—
Per 1 million tokens (USD)
Effort levels
3total
Default: Max
Benchmark results
32total
Benchmark types: 26

Model information

API model ID
kimi-k3
release date
Unknown
effort
Low / High / Max (default)
Effort source
Based on the HF model card for Kimi K3 (moonshotai/Kimi-K3): "Thinking effort is configured with the top-level reasoning_effort request field, which supports "low", "high", and "max" (default "max")." An official statement of the release date has not been confirmed.
Claude Code
unavailable
Codex
unavailable
Availability notes
Both are false because the model appears in neither the Claude Code official model list (code.claude.com/docs/en/model-config alias table, retrieved 2026-11-13; original: the table for "Use a model alias to select model settings without remembering exact version numbers" contains only Anthropic models such as fable, sonnet, opus, and haiku) nor the Codex official model list (learn.chatgpt.com/docs/models, retrieved 2026-11-13; original: "Recommended models" contains only the GPT-6 series) (connections via a custom provider are not counted as official options, and connections via a gateway are not counted as official options).
Availability source
https://code.claude.com/docs/en/model-config

Price (USD per 1 million tokens)

Prices do not vary by effort; token consumption does.

input—
output—
cache read—
cache write—

Not collected.

Benchmark results

BenchmarkeffortScoreSource cost (metric)measured/reported bysourceConditions and extraction method
Terminal-Bench 2.1Max (default)Max88.3%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0. Kimi Code harness.transcribed from table
Source excerptTerminal-Bench 2.1 | 88.3Location HF model card "3. Evaluation Results" table, Terminal-Bench 2.1 row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: Coding / Terminal-Bench 2.1
DeepSWE v1.1Max (default)Max67.5%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0. Kimi Code harness; DeepSWE v1.1 tasks (67.3 is separately stated for mini-SWE-agent).transcribed from table
Source excerptDeepSWE | 67.5Location HF model card "3. Evaluation Results" table, DeepSWE row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: Coding / DeepSWE
PostTrainBenchMax (default)Max36.6%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0. Official Harbor implementation; maximum effort; average of 3 runs on H20 GPU instead of H100; Claude Code harness.transcribed from table
Source excerptPostTrainBench | 36.6Location HF model card "3. Evaluation Results" table, PostTrainBench row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: Coding / PostTrainBench
CharXiv ReasoningMax (default)Max84.8%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0transcribed from table
Source excerptCharXiv (RQ) | 84.8 / 91.3Location HF model card "3. Evaluation Results" table, CharXiv (RQ) (left-hand value in the without tools / with tools pair) row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: Multimodal (average of 3 runs)
GPQA DiamondLowLow84.9%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, gpqa_diamond.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Kimi K3 (low)". effort is indicated by the suffix _low in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptgpqa_diamond.csv: Model version=kimi-k3_low, Best score (across scorers)=0.8484848484848485, stderr=0.025545650426603637, Started at=2026-08-07T00:54:16.000ZLocation gpqa_diamond.csv, data row 85 (file lines 138-140; multi-line record), id=mvNb3hNgSZ2gS6jHMtvCdk, Model version=kimi-k3_low
GPQA DiamondHighHigh91.9%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, gpqa_diamond.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Kimi K3 (high)". effort is indicated by the suffix _high in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptgpqa_diamond.csv: Model version=kimi-k3_high, Best score (across scorers)=0.9191919191919192, stderr=0.019417681889724515, Started at=2026-08-07T00:53:51.000ZLocation gpqa_diamond.csv, data row 86 (file lines 141-143; multi-line record), id=AoAD5uVtgVDnXmwA7STuno, Model version=kimi-k3_high
GPQA DiamondMax (default)Max93.1%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, gpqa_diamond.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Kimi K3 (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptgpqa_diamond.csv: Model version=kimi-k3_max, Best score (across scorers)=0.9311868686868687, stderr=0.014937225364194858, Started at=2026-07-16T21:59:40.000ZLocation gpqa_diamond.csv, data row 153 (file lines 217-219; multi-line record), id=B8MaPUe3EYwUZASsNqE7C9, Model version=kimi-k3_max
FrontierMath Tiers 1-3(v2)Max (default)Max72.2%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, frontiermath_tiers_1_3_v2.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Kimi K3 (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptfrontiermath_tiers_1_3_v2.csv: Model version=kimi-k3_max, Best score (across scorers)=0.721830985915493, stderr=0.026636607935112643, Started at=2026-07-17T01:48:42.000ZLocation frontiermath_tiers_1_3_v2.csv, data row 88 (file lines 116-118; multi-line record), id=bcntamvGsSH6QgETdoaSph, Model version=kimi-k3_max
FrontierMath Tier 4(v2)Max (default)Max39.0%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, frontiermath_tier_4_v2.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Kimi K3 (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptfrontiermath_tier_4_v2.csv: Model version=kimi-k3_max, Best score (across scorers)=0.3902439024390244, stderr=0.07712872341874095, Started at=2026-07-17T01:48:42.000ZLocation frontiermath_tier_4_v2.csv, data row 42 (file lines 77-79; multi-line record), id=gZERSAQwxBAevQzBsWfUie, Model version=kimi-k3_max
OTIS Mock AIME 2024-2025LowLow68.9%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, otis_mock_aime_2024_2025.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Kimi K3 (low)". effort is indicated by the suffix _low in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptotis_mock_aime_2024_2025.csv: Model version=kimi-k3_low, Best score (across scorers)=0.6888888888888889, stderr=0.06979205927323111, Started at=2026-08-07T00:46:06.000ZLocation otis_mock_aime_2024_2025.csv, data row 93 (file lines 153-155; multi-line record), id=frszScNxpHXXjSHawCt9Zb, Model version=kimi-k3_low
OTIS Mock AIME 2024-2025HighHigh93.3%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, otis_mock_aime_2024_2025.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Kimi K3 (high)". effort is indicated by the suffix _high in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptotis_mock_aime_2024_2025.csv: Model version=kimi-k3_high, Best score (across scorers)=0.9333333333333333, stderr=0.03760507165451775, Started at=2026-08-07T00:46:05.000ZLocation otis_mock_aime_2024_2025.csv, data row 94 (file lines 156-158; multi-line record), id=4E5xGsuRMwMBhAfC4F2G4N, Model version=kimi-k3_high
OTIS Mock AIME 2024-2025Max (default)Max97.2%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, otis_mock_aime_2024_2025.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Kimi K3 (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptotis_mock_aime_2024_2025.csv: Model version=kimi-k3_max, Best score (across scorers)=0.9722222222222222, stderr=0.011158359139033182, Started at=2026-07-16T21:59:40.000ZLocation otis_mock_aime_2024_2025.csv, data row 159 (file lines 228-230; multi-line record), id=A33om9MTYMC7xYeLv9X7Ha, Model version=kimi-k3_max
SimpleQA VerifiedMax (default)Max50.6%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, simpleqa_verified.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Kimi K3 (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptsimpleqa_verified.csv: Model version=kimi-k3_max, Best score (across scorers)=0.506, stderr=0.015818160898606836, Started at=2026-08-27T19:36:12.000ZLocation simpleqa_verified.csv, data row 44 (file lines 73-75; multi-line record), id=mZLWt2JSYvJyGhMvji7CT8, Model version=kimi-k3_max
Chess Puzzles(Epoch AI)LowLow18.0%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, chess_puzzles.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Kimi K3 (low)". effort is indicated by the suffix _low in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptchess_puzzles.csv: Model version=kimi-k3_low, Best score (across scorers)=0.18, stderr=0.03861229196653691, Started at=2026-10-13T17:06:54.000ZLocation chess_puzzles.csv, data row 32 (file lines 66-68; multi-line record), id=VPSX7EnwWuBrCPnpYGMUEM, Model version=kimi-k3_low
Chess Puzzles(Epoch AI)HighHigh37.0%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, chess_puzzles.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Kimi K3 (high)". effort is indicated by the suffix _high in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptchess_puzzles.csv: Model version=kimi-k3_high, Best score (across scorers)=0.37, stderr=0.048523658709390974, Started at=2026-10-13T17:10:13.000ZLocation chess_puzzles.csv, data row 31 (file lines 63-65; multi-line record), id=X2QNPHFXuWosA6AiuvVtDv, Model version=kimi-k3_high
Chess Puzzles(Epoch AI)Max (default)Max36.0%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, chess_puzzles.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Kimi K3 (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptchess_puzzles.csv: Model version=kimi-k3_max, Best score (across scorers)=0.36, stderr=0.048241815132442176, Started at=2026-10-13T17:13:35.000ZLocation chess_puzzles.csv, data row 30 (file lines 60-62; multi-line record), id=nzQV9YvictnDF2EwxqTJQs, Model version=kimi-k3_max
MirrorCodeMax (default)Max54.4%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, mirrorcode.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Kimi K3 (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptmirrorcode.csv: Model version=kimi-k3_max, Best score (across scorers)=0.5444444444444444, stderr=0.11439589045541113, Started at=2026-09-03T21:02:17.000ZLocation mirrorcode.csv, data row 37 (file lines 55-57; multi-line record), id=beEmcaWF8wxsWTLD5bMYxi, Model version=kimi-k3_max
Epoch Capabilities Index(ECI)not statednot stated158.58—Epoch AIEpoch AI, epoch_capabilities_index/eci_scores.csv (CC BY 4.0)third-party · Retrieved 2026-11-12A per-model metric that combines results across multiple benchmarks and multiple efforts (not per effort). Calculated by Epoch AI.transcribed from table
Source excerpteci_scores.csv: Model=Kimi K3, eci=158.58, eci_ci_low=156.6, eci_ci_high=160.8Location epoch_capabilities_index/eci_scores.csv, data row 24 (file line 25), Model=Kimi K3
SWE-Marathon (version not stated)Max (default)Max42.0%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0. Claude Code harness. Evaluated on an official task branch adjusted for H20 before the final version of v1.1 (as of 2026-07-09).transcribed from table
Source excerptSWE-Marathon | 42.0Location HF model card "3. Evaluation Results" table, SWE-Marathon row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: Coding / SWE-Marathon
MLS-Bench-LiteMax (default)Max48.3%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0. Kimi Code harness.transcribed from table
Source excerptMLS-Bench-Lite | 48.3Location HF model card "3. Evaluation Results" table, MLS-Bench-Lite row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: Coding / MLS-Bench-Lite
AutomationBench (600 public tasks)Max (default)Max30.8%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0. Subset of 600 public tasks; official GitHub configuration.transcribed from table
Source excerptAutomationBench | 30.8Location HF model card "3. Evaluation Results" table, AutomationBench row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: Agentic / AutomationBench
BrowseCompMax (default)Max91.2%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0. Context compression at 300K tokens (90.4 is separately stated for 1M without context management).transcribed from table
Source excerptBrowseComp | 91.2Location HF model card "3. Evaluation Results" table, BrowseComp row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: Agentic / BrowseComp
MCP AtlasMax (default)Max84.2%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0. Subset of 500 public tasks; limit of 100 turns; grading by Gemini 3.1 Pro.transcribed from table
Source excerptMCP-Atlas | 84.2Location HF model card "3. Evaluation Results" table, MCP-Atlas row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: Agentic / MCP-Atlas
CharXiv Reasoning (with tools)Max (default)Max91.3%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0transcribed from table
Source excerptCharXiv (RQ) | 84.8 / 91.3Location HF model card "3. Evaluation Results" table, CharXiv (RQ) (right-hand value in the without tools / with tools pair) row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: table note (vision tools use Python)
MMMU-ProMax (default)Max81.6%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0transcribed from table
Source excerptMMMU-Pro | 81.6 / 83.4Location HF model card "3. Evaluation Results" table, MMMU-Pro (left-hand value in the without tools / with tools pair) row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: Multimodal (average of 3 runs, official protocol)
MMMU-Pro (with tools)Max (default)Max83.4%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0transcribed from table
Source excerptMMMU-Pro | 81.6 / 83.4Location HF model card "3. Evaluation Results" table, MMMU-Pro (right-hand value in the without tools / with tools pair) row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: table note (vision tools use Python)
MathVisionMax (default)Max94.3%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0transcribed from table
Source excerptMathVision | 94.3 / 97.8Location HF model card "3. Evaluation Results" table, MathVision (left-hand value in the without tools / with tools pair) row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: Multimodal (average of 3 runs)
MathVision (with tools)Max (default)Max97.8%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0transcribed from table
Source excerptMathVision | 94.3 / 97.8Location HF model card "3. Evaluation Results" table, MathVision (right-hand value in the without tools / with tools pair) row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: table note (vision tools use Python)
BabyVision (with tools)Max (default)Max85.7%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0. With tools (Python).transcribed from table
Source excerptBabyVision w/ python | 85.7Location HF model card "3. Evaluation Results" table, BabyVision w/ python row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: Multimodal (average of 3 runs)
ZeroBench main (with tools, pass@5)Max (default)Max41.0%—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0. With tools (Python).transcribed from table
Source excerptZeroBench (pass@5) | 23.0 / 41.0Location HF model card "3. Evaluation Results" table, ZeroBench (pass@5) (right-hand value in the without tools / with tools pair) row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: Multimodal (5 runs for ZeroBench under the official settings)
OfficeQA ProMax (default)Max63.3—Moonshot AIKimi K3 (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table note: All Kimi K3 results are obtained with reasoning effort set to 'max' and temperature = 1.0. Claude Code harness; PDF contains only page images.transcribed from table
Source excerptOfficeQA Pro | 63.3Location HF model card "3. Evaluation Results" table, OfficeQA Pro row, Kimi K3 (max) column (leftmost numeric column). Supporting footnote: Agentic / OfficeQA Pro
τ²-bench banking_knowledge (v1.0.1)Max (default)Max37.1%—Sierraτ²-bench leaderboard submission kimi-k3_sierra_2026-08-04 (sierra-research/tau2-bench, MIT)third-party · Retrieved 2026-11-13τ²-bench leaderboard submission kimi-k3_sierra_2026-08-04 (submission_type standard, tau2_bench_version 1.0.1). Submitter "Sierra", model_organization "Moonshot AI": Run by Sierra, which operates τ²-bench (third-party measurement). Verified on the leaderboard (trajectories available, prompts not modified, no questions omitted). The value is pass^1 (%). reasoning_effort "max". User simulator gpt-5.2 (reasoning_effort low). Search configuration retrieval_config "alltools". Evaluation date 2026-07-24. Notes summary: 4 trial runs; seed 300.transcribed from table
Source excerptbanking_knowledge.pass_1=37.11340206185567Location In submissions/kimi-k3_sierra_2026-08-04/submission.json: results.banking_knowledge.pass_1 (line 17 of file). model_name "Kimi K3". methodology.notes (line 36)

Related rankings

References