Capability
80.5/100
field median 58.2
#5 of 221 ranked models
Kimi K3
Released Jul 16, 2026 — see all recent releases
Decision reading
Kimi K3 scores 80.5 out of 100 and ranks #5 of 221. This profile shows 44 source-displayable benchmark rows; its strongest eligible category is Multimodal & Grounded at #4. API pricing is $3 input and $15 output per million tokens, with cached input at $0.3.
Data as of August 19, 2026 · How the score is built
Strongest published evidence
Multimodal & Grounded ranks #4. Particularly strong for screenshots, documents, charts, and grounded multimodal workflows.
Validate before choosing
44 published rows leave some tracked benchmark slots empty. Knowledge is its lowest eligible category at #5.
Decision snapshot
Each value carries a field reference instead of floating alone. Markers compare this model with the current ranked and priced catalog; they are not absolute quality thresholds.
Price
$3input / $15 output
input median $1
cached $0.30 · blended $9
Speed
38tok/s
field median 97 tok/s
First token 55.01 s
Context
1.05Mtokens
field median 203,000
Reported for this model; direct source link not stored
Capability shape
Each axis shows percentile within that category’s eligible cohort. The comparison outline is the median of the six nearest public-score peers; a collapsed vertex means the category is not rank-eligible.
Kimi K3 category percentile values
- Agentic97th percentile
- Coding97th percentile
- ReasoningNot eligible
- Knowledge93rd percentile
- MathNot eligible
- MultilingualNot eligible
- Multimodal91st percentile
- Instruction followingNot eligible
The dashed outline is median of 6 nearest peers.
Eligible category ranks
- Agentic#5/134
- Coding#5/138
- ReasoningNot ranked
- Knowledge#5/57
- MathNot ranked
- MultilingualNot ranked
- Multimodal#4/33
- Inst. FollowingNot ranked
Top decileTop quartileMid-fieldNot eligible
What it costs to get this score
Published API price against the public score. The x-axis uses a log scale; the dashed path marks models that are not beaten by a cheaper, higher-scoring option. Price uses average of published input and output rates.
The chart opens on the current model. Scroll horizontally to inspect the full price axis.
Current modelKimi K3 · 80.5 score · $9 blended per million tokens
Horizontal: blended price per million tokens, log scale · Vertical: public score
How much of this is verified
Coverage is split by category so a strong number never hides a thin evidence base. Verified means the row is tied to a published source; provisional rows remain visible but separate.
- Agentic10/10 verified
- Coding11/13 verified
- ReasoningNot measured
- Knowledge4/4 verified
- MathNot measured
- MultilingualNot measured
- Multimodal13/13 verified
- Inst. FollowingNot measured
Verified sourceProvisionalNot measured
Spec sheet
Each documented value carries its source. Missing fields stay visible as not sourced or not published, rather than disappearing from the page.
- API model ID
- Not published
- Context window
- 1.05M
- Maximum output
- Not sourced yet
- Knowledge cutoff
- Not sourced yet
- Input modalities
- Not sourced yet
- Output modalities
- Not sourced yet
- Parameters
- Not sourced yet
- Availability
- Not sourced yet
- Cloud regions
- Not tracked yet
- Lifecycle
- Current
- API capabilities
- Tool calling, structured outputs, and batch support are not tracked yet
- Prompt caching
- Published at $0.30 per million cached input tokens
- Self-host
- Weight availability is pending
- Rate limits
- Not tracked yet
Deployment options
Self-host and provider-specific paths stay separate from benchmark evidence so operating constraints are visible before a score becomes the whole decision.
Category score record
Scores and ranks appear only where published evidence can be displayed. The table keeps the score, weight, cohort, and evidence state together.
| Category | Score | Rank | Percentile | Weight | Benchmarks | Evidence |
|---|---|---|---|---|---|---|
| AgenticRank #5 of 134Percentile 97thWeight 22%10 benchmarksVerified | 73.9 | #5 of 134 | 97th | 22% | 10 benchmarks | Verified |
| CodingRank #5 of 138Percentile 97thWeight 20%13 benchmarksMixed sources | 78.1 | #5 of 138 | 97th | 20% | 13 benchmarks | Mixed sources |
| ReasoningWeight 17%0 benchmarksNot measured | Not measured | Not ranked | Not available | 17% | 0 benchmarks | Not measured |
| KnowledgeRank #5 of 57Percentile 93rdWeight 12%4 benchmarksVerified | 84.8 | #5 of 57 | 93rd | 12% | 4 benchmarks | Verified |
| MathWeight 5%0 benchmarksNot measured | Not measured | Not ranked | Not available | 5% | 0 benchmarks | Not measured |
| MultilingualWeight 7%0 benchmarksNot measured | Not measured | Not ranked | Not available | 7% | 0 benchmarks | Not measured |
| MultimodalRank #4 of 33Percentile 91stWeight 12%13 benchmarksVerified | 87.9 | #4 of 33 | 91st | 12% | 13 benchmarks | Verified |
| Inst. FollowingWeight 5%0 benchmarksNot measured | Not measured | Not ranked | Not available | 5% | 0 benchmarks | Not measured |
Benchmark ledger
Coding opens by default. The marker compares each value with the best source-verified result in the catalog; provisional leaders do not set the reference. Expand the remaining categories for every published row.
Coding13 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| deepSwe | Score67.5% | Versus best verified row Best verified: GPT-5.6 Sol · 72.7% | Gap5.2 behind | WeightDisplay only | Provider exact |
| cursorBench32 | Score60.8% | Versus best verified row Best verified: Grok 4.6 · 70.8% | Gap10 behind | WeightDisplay only | Benchmark exact |
| FrontierSWE | Score81.2% | Versus best verified row Best verified: Kimi K3 · 81.2% | GapBest verified | WeightDisplay only | Provider exact |
| ProgramBenchProgramBench: Can Language Models Rebuild Programs From Scratch? | Score77.8% | Versus best verified row Best verified: Claude Opus 5 · 93.0% | Gap15.2 behind | WeightDisplay only | Provider exact |
| Kimi Code Bench v2 | Score72.9% | Versus best verified row Best verified: Kimi K3 · 72.9% | GapBest verified | WeightDisplay only | Provider exact |
| sweMarathon | Score42% | Versus best verified row Best verified: GLM-5.3 · 42.5% | Gap0.5 behind | WeightDisplay only | Provider exact |
| PostTrain Bench | Score36.6% | Versus best verified row Best verified: GLM-5.3 · 39.8% | Gap3.2 behind | WeightDisplay only | Provider exact |
| MLS-Bench Lite | Score48.3% | Versus best verified row Best verified: Kimi K3 · 48.3% | GapBest verified | WeightDisplay only | Provider exact |
| VulcanBench v3 | Score73.9% | Versus best verified row Best verified: Grok 4.5 · 91.3% | Gap17.4 behind | WeightDisplay only | Benchmark exact |
| APEX-SWE | Score48.0% | Versus best verified row Best verified: Claude Fable 5 · 58.8% | Gap10.8 behind | WeightDisplay only | Benchmark exact |
| EEBenchEEBench V1 core corpus | Score38.3% | Versus best verified row Best verified: Grok 4.6 · 58.3% | Gap20 behind | WeightDisplay only | Benchmark exact |
| InferenceEval | Score39.8% | Versus best verified row Best verified: Grok 4.6 · 46.9% | Gap7.1 behind | WeightDisplay only | Secondary exact |
| KernelBench Internal | Score68.5% | Versus best verified row Best verified: Grok 4.6 · 70.6% | Gap2.1 behind | WeightDisplay only | Secondary exact |
Agentic10 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| Terminal-Bench 2.0 | Score88.3% | Versus best verified row Best verified: GPT-5.6 Sol · 91.9% | Gap3.6 behind | WeightWeighted 38% | Provider exact |
| BrowseComp | Score91.2% | Versus best verified row Best verified: GPT-5.6 Sol · 92.2% | Gap1 behind | WeightWeighted 28% | Provider exact |
| DeepSearchQA | Score95.0% | Versus best verified row Best verified: Claude Opus 5 · 95.0% | GapBest verified | WeightDisplay only | Provider exact |
| Toolathlon-Verified | Score73.2% | Versus best verified row Best verified: Claude Opus 5 · 80.6% | Gap7.4 behind | WeightDisplay only | Provider exact |
| MCP Atlas | Score84.2% | Versus best verified row Best verified: Muse Spark 1.1 · 88.1% | Gap3.9 behind | WeightDisplay only | Provider exact |
| AutomationBench | Score30.8% | Versus best verified row Best verified: GLM-5.3 · 48.2% | Gap17.4 behind | WeightDisplay only | Provider exact |
| JobBench | Score52.9% | Versus best verified row Best verified: Muse Spark 1.1 · 54.7% | Gap1.8 behind | WeightDisplay only | Provider exact |
| APEX-Agents | Score37.6% | Versus best verified row Best verified: Grok 4.6 · 57.5% | Gap19.9 behind | WeightDisplay only | Provider exact |
| SpreadsheetBench 2 | Score34.8% | Versus best verified row Best verified: Kimi K3 · 34.8% | GapBest verified | WeightDisplay only | Provider exact |
| DECK-BenchDECK-Bench (Internal) | Score73.5% | Versus best verified row Best verified: Kimi K3 · 73.5% | GapBest verified | WeightDisplay only | Provider exact |
Knowledge4 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| HLEHumanity's Last Exam | Score56% | Versus best verified row Best verified: Claude Opus 5 · 64.7% | Gap8.7 behind | WeightWeighted 45% | Provider exact |
| GPQAGraduate-Level Google-Proof Q&A | Score93.5% | Versus best verified row Best verified: Sakana Fugu-Ultra · 95.5% | Gap2 behind | WeightWeighted 7% | Provider exact |
| GPQA-DGPQA Diamond | Score93.5% | Versus best verified row Best verified: Sakana Fugu-Ultra · 95.5% | Gap2 behind | WeightDisplay only | Provider exact |
| HLE w/o toolsHumanity's Last Exam without tools | Score43.5% | Versus best verified row Best verified: Claude Mythos 5 · 59% | Gap15.5 behind | WeightDisplay only | Provider exact |
Multimodal13 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| MMMU-ProMassive Multi-discipline Multimodal Understanding Pro | Score81.6% | Versus best verified row Best verified: GPT-5.4 Pro · 94% | Gap12.4 behind | WeightWeighted 45% | Provider exact |
| OfficeQA Pro | Score63.3% | Versus best verified row Best verified: Claude Opus 5 · 66.9% | Gap3.6 behind | WeightWeighted 30% | Provider exact |
| CharXivCharXiv Reasoning | Score91.3% | Versus best verified row Best verified: Claude Mythos 5 · 93.5% | Gap2.2 behind | WeightWeighted 25% | Provider exact |
| MMMU-Pro w/ PythonMMMU-Pro with Python | Score83.4% | Versus best verified row Best verified: GPT-5.6 Sol · 84.6% | Gap1.2 behind | WeightDisplay only | Provider exact |
| CharXiv w/o toolsCharXiv Reasoning without tools | Score84.8% | Versus best verified row Best verified: Claude Mythos 5 · 88.9% | Gap4.1 behind | WeightDisplay only | Provider exact |
| MathVision | Score94.3% | Versus best verified row Best verified: Qwen3.8 Max · 95.2% | Gap0.9 behind | WeightDisplay only | Provider exact |
| MathVision w/ PythonMathVision with Python | Score97.8% | Versus best verified row Best verified: Kimi K3 · 97.8% | GapBest verified | WeightDisplay only | Provider exact |
| BabyVision w/ PythonBabyVision with Python | Score85.7% | Versus best verified row Best verified: Qwen3.8 Max · 91.3% | Gap5.6 behind | WeightDisplay only | Provider exact |
| ZeroBench | Score23.0% | Versus best verified row Best verified: Muse Spark · 33.0% | Gap10 behind | WeightDisplay only | Provider exact |
| ZeroBench w/ PythonZeroBench_main with Python | Score41.0% | Versus best verified row Best verified: Qwen3.8 Max · 49.0% | Gap8 behind | WeightDisplay only | Provider exact |
| WorldVQA ForceAnswer | Score51.0% | Versus best verified row Best verified: Kimi K3 · 51.0% | GapBest verified | WeightDisplay only | Provider exact |
| OmniDocBench | Score91.1% | Versus best verified row Best verified: Kimi K3 · 91.1% | GapBest verified | WeightDisplay only | Provider exact |
| PerceptionBenchPerceptionBench (Internal) | Score58.5% | Versus best verified row Best verified: Qwen3.8 Max · 63.5% | Gap5 behind | WeightDisplay only | Provider exact |
External signals4 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| ExploitBenchExploitBench v8-bench | Score32% | Versus best verified row Best verified: Claude Mythos 5 · 78% | Gap46 behind | WeightDisplay only | Benchmark exact |
| ACE solvedACE Cyber Range Challenges Solved | Score0% | Versus best verified row Best verified: Kimi K3 · 0% | GapBest verified | WeightDisplay only | Benchmark exact |
| The Last Ones stepsThe Last Ones Average Progress | Score17% | Versus best verified row Best verified: Kimi K3 · 17% | GapBest verified | WeightDisplay only | Benchmark exact |
| The Last Ones completionThe Last Ones Cyber Range Completion Rate | Score10% | Versus best verified row Best verified: GPT-5.6 Sol · 70% | Gap60 behind | WeightDisplay only | Benchmark exact |
Lineage
The sequence follows explicit supersedes links. Scores and prices remain blank when the corresponding public row or first-party rate is unavailable.
Feb 1, 2026
Kimi K2.5Score 58.9 · $0.6 / $3
Apr 20, 2026
Kimi K2.6Score 60.3 · $0.95 / $4
Jul 16, 2026 · you are here
Kimi K3Score 80.5 · $3 / $15
Base entry
How to read this profile
The visual layer above carries the decisions. These notes preserve the model, ranking, coverage, and family context behind the numbers.
Kimi K3 ranks #5 of 221 on the public leaderboard with a score of 80.53/100. Its source-verified position is #5 of 104.
Kimi K3 has no verified publication yet for weight access. Its 1.05M context window is documented separately from that pending weight status. The profile records its reasoning mode as reasoning.
Its explicit predecessor is Kimi K2.6. 44 of 437 tracked benchmark slots currently have displayable evidence. Missing categories stay blank.
Its strongest eligible category is Multimodal & Grounded at #4, while its lowest eligible position is Knowledge at #5. particularly strong for screenshots, documents, charts, and grounded multimodal workflows.
Radar
Full release historyKimi K3 release history
Kimi K3 released
Moonshot AI · Model release
Source confirmed
Frequently asked questions
How does Kimi K3 perform overall in AI benchmarks?
Kimi K3 ranks #5 out of 221 models on the public BenchAlign leaderboard, with a score of 80.53/100. Its evidence status is Supported, and this profile shows 44 source-displayable benchmark rows. The label describes evidence depth, not a provider quality claim; inspect category rows before choosing a workload.
Is Kimi K3 good for knowledge and understanding?
Kimi K3 ranks #5 out of 57 eligible models for knowledge and understanding, with a public category score of 84.8/100. That places it in the current top ten for this category. Check the underlying rows before treating the aggregate as a workload guarantee.
Is Kimi K3 good for coding and programming?
Kimi K3 ranks #5 out of 138 eligible models for coding and programming, with a public category score of 78.1/100. That places it in the current top ten for this category. Check the underlying rows before treating the aggregate as a workload guarantee.
Is Kimi K3 good for agentic tool use and computer tasks?
Kimi K3 ranks #5 out of 134 eligible models for agentic tool use and computer tasks, with a public category score of 73.9/100. That places it in the current top ten for this category. Check the underlying rows before treating the aggregate as a workload guarantee.
Is Kimi K3 good for multimodal and grounded tasks?
Kimi K3 ranks #4 out of 33 eligible models for multimodal and grounded tasks, with a public category score of 87.9/100. That places it in the current top ten for this category. Check the underlying rows before treating the aggregate as a workload guarantee.
Does Kimi K3 have full benchmark coverage on BenchLM?
No. Kimi K3 currently has 68 source-displayable rows across 437 tracked benchmark slots. The profile exposes published, non-generated evidence and leaves missing categories blank until an exact evaluation is available. Coverage describes how much was measured; it is not a penalty added to an individual benchmark result.
What is the context window size of Kimi K3?
Kimi K3 has a reported context window of 1.05M in the exact-model catalog record. The value stays visible, but the profile marks its source link as unavailable instead of presenting it as directly documented. Maximum output length remains separate because providers often publish a different limit.
Last updated August 19, 2026. Runtime fields remain blank until a sourced snapshot exists.
Watch Kimi K3 in the weekly brief
Get one weekly email when material rank, price, availability, or benchmark evidence changes are worth revisiting.
Read a sample issueJoin 2,000+ readers.