modelbench日本語

Kimi K3.1 Pro

Moonshot AICurrent

Input price
—
Per 1 million tokens (USD)
Output price
—
Per 1 million tokens (USD)
Effort levels
3total
Default: Max
Benchmark results
28total
Benchmark types: 28

Model information

API model ID
kimi-k3.1-pro
release date
Unknown
effort
Low / High / Max (default)
Effort source
Based on Kimi API Reasoning Effort documentation (platform.kimi.ai/docs/guide/use-thinking-effort) and 5. Model Usage in the HF model card (moonshotai/Kimi-K3.1-Pro): low, high, max, default is max.
Claude Code
unavailable
Codex
unavailable
Availability notes
Kimi Code is the official route for use. Claude Code is mentioned only as a benchmark harness. The official model list for Claude Code / Codex (checked 2026-11-12) contains only Anthropic / OpenAI models, respectively, so both are false.
Availability source
https://kimi.com/en/blog/kimi-k3-1-pro

Price (USD per 1 million tokens)

Prices do not vary by effort; token consumption does.

input—
output—
cache read—
cache write—

The figures in the official pricing table could not be read when the page was fetched, so they could not be collected. batch is 60% of standard pricing; K3 series cache write is charged separately by TTL (5min/1h). context 1,048,576

Source: https://platform.kimi.ai/docs/pricing/chat

Benchmark results

BenchmarkeffortScoreSource cost (metric)measured/reported bysourceConditions and extraction method
Terminal-Bench 4.0Max (default)Max65.7%—Moonshot AIIntroducing Kimi K3.1 Proofficial · Retrieved 2026-11-12Vendor-reported value. Self-reported by Moonshot. footnote: reasoning effort max, temperature 1.0, top-p 0.95. Evaluated with the Claude-Code harnesstranscribed from text
Source excerptTerminal-Bench 4.0 65.7%
Terminal-Bench-Science 0.1Max (default)Max38.1%—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95. Evaluated with Claude Code.transcribed from table
Source excerptTerminal-Bench-Science 0.1 | 38.1Location HF model card "3. Evaluation Results" table, Terminal-Bench-Science 0.1 row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Coding 2
DeepSWE v1.1Max (default)Max75.5%—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95. Evaluated with Claude Code.transcribed from table
Source excerptDeepSWE v1.1 | 75.5Location HF model card "3. Evaluation Results" table, DeepSWE v1.1 row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Coding 5
Chartography (with tools)Max (default)Max90.2%—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95. With tools (Python); average of 5 runs.transcribed from table
Source excerptChartoGraphy w/ python | 90.2Location HF model card "3. Evaluation Results" table, ChartoGraphy w/ python row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Agentic 11
BenchCAD (with tools, version not stated)Max (default)Max92.7—Moonshot AIIntroducing Kimi K3.1 Proofficial · Retrieved 2026-11-12Vendor-reported value. Self-reported by Moonshot. footnote: reasoning effort max, temperature 1.0, top-p 0.95. with toolstranscribed from text
Source excerptBenchCAD, with tools 92.7
BenchCAD (without tools, version not stated)Max (default)Max72.8—Moonshot AIIntroducing Kimi K3.1 Proofficial · Retrieved 2026-11-12Vendor-reported value. Self-reported by Moonshot. footnote: reasoning effort max, temperature 1.0, top-p 0.95. without toolstranscribed from text
Source excerptBenchCAD, without tools 72.8
ExploitGymMax (default)Max20.7%—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95. Kimi Code; 6 hours per 1 task.transcribed from table
Source excerptExploitGym | 20.7Location HF model card "3. Evaluation Results" table, ExploitGym row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Cyber 2
GDP.PDFMax (default)Max34.6%—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95. Strict pass rate for answers with all items graded; grading by Gemini-3.5-Flash Medium.transcribed from table
Source excerptGDP.pdf | 34.6Location HF model card "3. Evaluation Results" table, GDP.pdf row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Reasoning 1
CharXiv ReasoningMax (default)Max89.0%—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95transcribed from table
Source excerptCharXiv (RQ) | 89.0Location HF model card "3. Evaluation Results" table, CharXiv (RQ) row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Reasoning 3 (average of 4 runs)
OfficeQA Pro v2Max (default)Max64.4—Moonshot AIIntroducing Kimi K3.1 Proofficial · Retrieved 2026-11-12Vendor-reported value. Self-reported by Moonshot. footnote: reasoning effort max, temperature 1.0, top-p 0.95.transcribed from text
Source excerptOfficeQA Pro v2 64.4
Frontier FinanceMax (default)Max58.6—Moonshot AIIntroducing Kimi K3.1 Proofficial · Retrieved 2026-11-12Vendor-reported value. Self-reported by Moonshot. footnote: reasoning effort max, temperature 1.0, top-p 0.95.transcribed from text
Source excerptFrontier Finance 58.6
GPQA DiamondMax (default)Max94.2%—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 K3.1 Pro (Together, max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12. Epoch ran it via Together's inference API (Model version kimi-k3.1-pro-together).transcribed from table
Source excerptgpqa_diamond.csv: Model version=kimi-k3.1-pro-together_max, Best score (across scorers)=0.9419191919191919, stderr=0.01439446786793499, Started at=2026-11-06T00:59:23.000ZLocation gpqa_diamond.csv, data row 2 (file line 3), id=2bhYPjF6VRWc5CeERGq8g5, Model version=kimi-k3.1-pro-together_max
FrontierMath Tiers 1-3(v2)Max (default)Max78.3%—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 K3.1 Pro (Together, max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12. Epoch ran it via Together's inference API (Model version kimi-k3.1-pro-together).transcribed from table
Source excerptfrontiermath_tiers_1_3_v2.csv: Model version=kimi-k3.1-pro-together_max, Best score (across scorers)=0.7824561403508772, stderr=0.024481850763866792, Started at=2026-11-06T00:59:23.000ZLocation frontiermath_tiers_1_3_v2.csv, data row 2 (file line 3), id=9iqojBnRuywLNkMDSVdwpJ, Model version=kimi-k3.1-pro-together_max
FrontierMath Tier 4(v2)Max (default)Max65.9%—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 K3.1 Pro (Together, max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12. Epoch ran it via Together's inference API (Model version kimi-k3.1-pro-together).transcribed from table
Source excerptfrontiermath_tier_4_v2.csv: Model version=kimi-k3.1-pro-together_max, Best score (across scorers)=0.6585365853658537, stderr=0.07497768853141171, Started at=2026-11-06T00:59:23.000ZLocation frontiermath_tier_4_v2.csv, data row 2 (file line 3), id=o7TxEhAxnN2qKUqqhFgp5H, Model version=kimi-k3.1-pro-together_max
OTIS Mock AIME 2024-2025Max (default)Max100.0%—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 K3.1 Pro (Together, max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12. Epoch ran it via Together's inference API (Model version kimi-k3.1-pro-together).transcribed from table
Source excerptotis_mock_aime_2024_2025.csv: Model version=kimi-k3.1-pro-together_max, Best score (across scorers)=1.0, stderr=0.0, Started at=2026-11-06T00:59:23.000ZLocation otis_mock_aime_2024_2025.csv, data row 2 (file line 3), id=8Qe2qf4Eg77PMfp2Eh7V6K, Model version=kimi-k3.1-pro-together_max
SimpleQA VerifiedMax (default)Max62.2%—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 K3.1 Pro (Together, max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12. Epoch ran it via Together's inference API (Model version kimi-k3.1-pro-together).transcribed from table
Source excerptsimpleqa_verified.csv: Model version=kimi-k3.1-pro-together_max, Best score (across scorers)=0.622, stderr=0.015341165254026547, Started at=2026-11-06T00:59:23.000ZLocation simpleqa_verified.csv, data row 2 (file line 3), id=HumfsXVh5BZWuTo6Z8Gqnf, Model version=kimi-k3.1-pro-together_max
Chess Puzzles(Epoch AI)Max (default)Max45.0%—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 K3.1 Pro (Together, max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12. Epoch ran it via Together's inference API (Model version kimi-k3.1-pro-together).transcribed from table
Source excerptchess_puzzles.csv: Model version=kimi-k3.1-pro-together_max, Best score (across scorers)=0.45, stderr=0.05, Started at=2026-11-06T00:59:23.000ZLocation chess_puzzles.csv, data row 2 (file line 3), id=8eSYUWwajdBjKiiM9CwnKu, Model version=kimi-k3.1-pro-together_max
Epoch Capabilities Index(ECI)not statednot stated161.37—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 K3.1 Pro, eci=161.37, eci_ci_low=158.99, eci_ci_high=164.0Location epoch_capabilities_index/eci_scores.csv, data row 16 (file line 17), Model=Kimi K3.1 Pro
SWE-Marathon v1.1Max (default)Max51.7%—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95. Evaluated with Claude Code.transcribed from table
Source excerptSWE Marathon v1.1 | 51.7Location HF model card "3. Evaluation Results" table, SWE Marathon v1.1 row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Coding 3
ProgramBench(Almost Solved)Max (default)Max31.5%—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95. Binary aggregation assigning 1 when a task's ProgramBench score is at least 95%; evaluated with Claude Code.transcribed from table
Source excerptProgram Bench | 31.5Location HF model card "3. Evaluation Results" table, Program Bench row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Coding 4
BrowseCompMax (default)Max92.1%—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95. Context compression at 200K tokens; maximum 500 steps; top-p 1.0.transcribed from table
Source excerptBrowseComp | 92.1Location HF model card "3. Evaluation Results" table, BrowseComp row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Agentic 9
CharXiv Reasoning (with tools)Max (default)Max93.4%—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95. With tools (Python); average of 4 runs.transcribed from table
Source excerptCharXiv (RQ) w/ python | 93.4Location HF model card "3. Evaluation Results" table, CharXiv (RQ) w/ python row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Agentic 11
MMMU-ProMax (default)Max84.1%—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95. Official protocol (preserving input order and placing images first).transcribed from table
Source excerptMMMU-Pro | 84.1Location HF model card "3. Evaluation Results" table, MMMU-Pro row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Reasoning 4
MMMU-Pro (with tools)Max (default)Max90.0%—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95. With tools (Python); average of 4 runs.transcribed from table
Source excerptMMMU-Pro w/ python | 90.0Location HF model card "3. Evaluation Results" table, MMMU-Pro w/ python row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Agentic 11
BabyVision (with tools)Max (default)Max95.4%—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95. With tools (Python); average of 4 runs.transcribed from table
Source excerptBabyVision w/ python | 95.4Location HF model card "3. Evaluation Results" table, BabyVision w/ python row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Agentic 11
ZeroBench main (with tools, pass@5)Max (default)Max63.0%—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95. With tools (Python); 5 runs.transcribed from table
Source excerptZeroBench_main w/ python (pass@5) | 63.0Location HF model card "3. Evaluation Results" table, ZeroBench_main w/ python (pass@5) row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Agentic 11
OfficeQA ProMax (default)Max67.4—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95. Evaluated with Claude Code; PDF contains only page images.transcribed from table
Source excerptOfficeQA Pro | 67.4Location HF model card "3. Evaluation Results" table, OfficeQA Pro row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Agentic 3
PostTrainBench v1.2Max (default)Max36.5%—Moonshot AIKimi K3.1 Pro (Hugging Face model card, moonshotai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Moonshot AI). Table column heading Kimi K3.1 Pro (max), table note: all Kimi K3.1 Pro results use reasoning effort max, temperature 1.0, top-p 0.95. v1.2 evaluated in-house using H20 GPU instead of H100, with Claude Code.transcribed from table
Source excerptPostTrain Bench v1.2 | 36.5Location HF model card "3. Evaluation Results" table, PostTrain Bench v1.2 row, Kimi K3.1 Pro (max) column (leftmost numeric column). Supporting footnote: Coding 6

Related rankings

References