Kimi 2.6 released
Moonshot AI · Model release
Source confirmed
Data as of October 10, 2026 · How the score is built
SupersededOpen WeightReasoning
Released Apr 20, 2026256K context
Kimi K2.6
Decision readingKimi K2.6 scores 58.7 out of 100 and ranks #53 of 218. This profile shows 37 source-displayable benchmark rows; its strongest eligible category is Mathematics at #1. API pricing is $0.95 input and $4 output per million tokens, with cached input at $0.16.
Released Apr 20, 2026 — see all recent releases
Kimi K2.6 will be repriced, updated or retired. Get each notice with its source and date. Follow model changes
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.
Capability
58.7/100
field median 50.1#53 of 218 ranked models
Public
#53of 218
Verified —
Price
$0.95input / $4 output
input median $0.95cached $0.16 · blended $2.48
Speed
60tok/s
field median 97 tok/sFirst token 2.67 s
Context
256Ktokens
field median 256,000Reported for this model; direct source link not stored
Strongest published evidence
Mathematics ranks #1. Particularly strong for mathematical reasoning, scientific computing, and quantitative analysis.
Validate before choosing
37 published rows leave some tracked benchmark slots empty. Agentic is its lowest eligible category at #57.
Source-linked · 37 displayable benchmark rows
Follow model changesCategory 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 #57 of 123Percentile 54thWeight 22%12 benchmarksVerified | 43.7 | 12 benchmarks | Verified | |||
| CodingRank #54 of 146Percentile 63rdWeight 20%10 benchmarksVerified | 43.9 | 10 benchmarks | Verified | |||
| ReasoningWeight 17%0 benchmarksNot measured | Not measured | 0 benchmarks | Not measured | |||
| MultimodalRank #27 of 54Percentile 51stWeight 12%5 benchmarksVerified | 73.8 | 5 benchmarks | Verified | |||
| KnowledgeRank #45 of 177Percentile 75thWeight 12%5 benchmarksVerified | 59.8 | 5 benchmarks | Verified | |||
| MultilingualWeight 7%0 benchmarksNot measured | Not measured | 0 benchmarks | Not measured | |||
| Inst. FollowingWeight 5%0 benchmarksNot measured | Not measured | 0 benchmarks | Not measured | |||
| MathRank #1 of 7Percentile 100thWeight 5%5 benchmarksVerified | 72.2 | 5 benchmarks | Verified |
37 of 667 tracked benchmark slots have displayable evidence · bars run 0–100
Coverage detailsHow 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.
- Agentic12/12 verified
- Coding10/10 verified
- ReasoningNot measured
- Multimodal5/5 verified
- Knowledge5/5 verified
- MultilingualNot measured
- Inst. FollowingNot measured
- Math5/5 verified
Verified sourceProvisionalNot measured
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 K2.6 category percentile values
- Agentic54th percentile
- Coding63rd percentile
- ReasoningNot eligible
- Multimodal51st percentile
- Knowledge75th percentile
- MultilingualNot eligible
- Instruction followingNot eligible
- Math100th percentile
The dashed outline is median of 6 nearest peers.
Eligible category ranks
- Agentic#57/123
- Coding#54/146
- ReasoningNot ranked
- Multimodal#27/54
- Knowledge#45/177
- MultilingualNot ranked
- Inst. FollowingNot ranked
- Math#1/7
Top decileTop quartileMid-fieldNot eligible
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.
Coding10 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| SWE-bench Pro | Score58.6% | Versus best verified row Best verified: Claude Opus 5.5 · 89.9% | Gap31.3 behind | Weight26% ref. weight | Provider exact |
| SciCodeScientific Code Benchmark | Score52.2% | Versus best verified row Best verified: Sakana Fugu · 60.1% | Gap7.9 behind | Weight10% ref. weight | Provider exact |
| LiveCodeBench (Vals)LiveCodeBench, Vals AI run | Score86.8% | Versus best verified row Best verified: Claude Fable 5.1 · 90.5% | Gap3.7 behind | Weight8% ref. weight | |
| SWE Multilingual | Score76.7% | Versus best verified row Best verified: Claude Opus 5.5 · 93.9% | Gap17.2 behind | Weight5% ref. weight | Provider exact |
| Terminal-Bench 2.0 | Score66.7% | Versus best verified row Best verified: GPT-5.5 · 82.0% | Gap15.3 behind | WeightScored in Agentic | Provider exact |
| SWE-bench VerifiedSoftware Engineering Benchmark Verified | Score80.2% | Versus best verified row Best verified: Claude Opus 5 · 96% | Gap15.8 behind | WeightDisplay only | Provider exact |
| LiveCodeBench v6 | Score89.6% | Versus best verified row Best verified: Sakana Fugu-Ultra · 93.2% | Gap3.6 behind | WeightDisplay only | Provider exact |
| Vibe Code BenchVibe Code Bench v1.1 | Score37.89% | Versus best verified row Best verified: Gemini 4 Argon · 91.90% | Gap54 behind | WeightDisplay only | Benchmark exact |
| cursorBench31 | Score47.6% | Versus best verified row Best verified: Claude Fable 5 · 70.6% | Gap23 behind | WeightDisplay only | Benchmark exact |
| SWE-bench (Vals)SWE-bench, Vals AI run | Score76.2% | Versus best verified row Best verified: Claude Opus 5 · 97.0% | Gap20.8 behind | WeightDisplay only | Verified |
Agentic12 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| OSWorld 2.0 | Score4.6% | Versus best verified row Best verified: GPT-6 Astra · 72.6% | Gap68 behind | Weight10% ref. weight | Benchmark exact |
| BrowseComp | Score83.2% | Versus best verified row Best verified: Atria Dawn Preview · 92.5% | Gap9.3 behind | Weight8% ref. weight | Provider exact |
| Terminal-Bench 2.0 | Score66.7% | Versus best verified row Best verified: GPT-5.5 · 82% | Gap15.3 behind | Weight6% ref. weight | Provider exact |
| OSWorld-Verified | Score73.1% | Versus best verified row Best verified: Qwen3.8 Max · 86.1% | Gap13 behind | Weight6% ref. weight | Provider exact |
| MCP Atlas | Score55.9% | Versus best verified row Best verified: Muse Spark 1.1 · 88.1% | Gap32.2 behind | Weight4% ref. weight | Provider exact |
| Terminal-Bench 2.1 (Vals)Terminal-Bench 2.1, Vals AI run | Score53.6% | Versus best verified row Best verified: GPT-6 Astra · 87.3% | Gap33.7 behind | Weight3% ref. weight | |
| Toolathlon | Score50% | Versus best verified row Best verified: Muse Spark 1.1 · 75.6% | Gap25.6 behind | Weight3% ref. weight | Provider exact |
| DeepSearchQA | Score92.5% | Versus best verified row Best verified: Atria Dawn Preview · 96.0% | Gap3.5 behind | Weight2% ref. weight | Provider exact |
| Claw-Eval | Score62.3% | Versus best verified row Best verified: Ornith-1.5-397B · 81.4% | Gap19.1 behind | WeightDisplay only | Benchmark exact |
| WideResearch | Score80.8% | Versus best verified row Best verified: Hy4 preview · 83.9% | Gap3.1 behind | WeightDisplay only | Provider exact |
| Gert LabsGert Labs Composite Game Benchmark | Score56.82% | Versus best verified row Best verified: Claude Opus 4.8 · 72.97% | Gap16.2 behind | WeightDisplay only | Benchmark exact |
| ResearchClawBench | Score18.0% | Versus best verified row Best verified: Claude Opus 4.8 · 21.1% | Gap3.1 behind | WeightDisplay only | Benchmark exact |
Multimodal5 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| MMMU-ProMassive Multi-discipline Multimodal Understanding Pro | Score79.4% | Versus best verified row Best verified: Gemini 3.5 Flash · 83.6% | Gap4.2 behind | WeightWeighted 40% | Provider exact |
| CharXivCharXiv Reasoning | Score80.4% | Versus best verified row Best verified: Qwen3.8 Max · 93.5% | Gap13.1 behind | WeightWeighted 20% | Provider exact |
| MMMU-Pro w/ PythonMMMU-Pro with Python | Score80.1% | Versus best verified row Best verified: GPT-5.6 Sol · 84.6% | Gap4.5 behind | WeightDisplay only | Provider exact |
| MathVision | Score87.4% | Versus best verified row Best verified: Qwen3.8 Max · 95.2% | Gap7.8 behind | WeightDisplay only | Provider exact |
| V* | Score96.9% | Versus best verified row Best verified: Qwen3.6 Plus · 96.9% | GapBest verified | WeightDisplay only | Provider exact |
Knowledge5 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| HLEHumanity's Last Exam | Score34.7% | Versus best verified row Best verified: Claude Fable 5.1 · 65% | Gap30.3 behind | Weight44% ref. weight | Provider exact |
| MMLU-Pro (Vals)MMLU-Pro, Vals AI run | Score87.6% | Versus best verified row Best verified: Claude Fable 5.1 · 92.4% | Gap4.8 behind | Weight6% ref. weight | Verified |
| GPQAGraduate-Level Google-Proof Q&A | Score90.5% | Versus best verified row Best verified: GPT-6 Astra · 96% | Gap5.5 behind | Weight3% ref. weight | Provider exact |
| GPQA Diamond (Vals)GPQA Diamond, Vals AI run | Score89.1% | Versus best verified row Best verified: Gemini 3.1 Pro · 95.5% | Gap6.4 behind | Weight2% ref. weight | |
| GPQA-DGPQA Diamond | Score90.5% | Versus best verified row Best verified: GPT-6 Astra · 96.0% | Gap5.5 behind | WeightDisplay only | Provider exact |
Math5 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| FrontierMath v2 (Tiers 1-3)FrontierMath v2 Tiers 1-3 | Score38.966% | Versus best verified row Best verified: GPT-5.6 Sol · 89.000% | Gap50 behind | WeightWeighted 30% | Benchmark exact |
| AIME26AIME 2026 | Score96.4% | Versus best verified row Best verified: GLM-5.2 · 99.2% | Gap2.8 behind | WeightWeighted 25% | Provider exact |
| HMMT Feb 2026Harvard-MIT Mathematics Tournament February 2026 | Score92.7% | Versus best verified row Best verified: Qwen3.7 Max · 97.1% | Gap4.4 behind | WeightWeighted 25% | Provider exact |
| FrontierMath v2 (Tier 4)FrontierMath v2 Tier 4 | Score14.580% | Versus best verified row Best verified: GPT-6 Astra · 97.600% | Gap83 behind | WeightWeighted 10% | Benchmark exact |
| MMAnswerBench | Score86.0% | Versus best verified row Best verified: GLM-5.2 · 91.0% | Gap5 behind | WeightDisplay only | Provider exact |
Bars run 0–100; the dark tick marks the best source-verified value
All 37 rowsWhat 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.
Current modelExplore all models
Kimi K2.6 · 58.7 score · $2.48 blended per million tokens
The chart opens on the current model. Scroll horizontally to inspect the full price axis.
Horizontal: blended price per million tokens, log scale · Vertical: public score
Lineage
The sequence follows explicit supersedes links. Each score is estimated for that model; a relative can inform a sparse estimate but never sets a floor, so a newer release can score below an earlier one. Scores and prices remain blank when the corresponding public row or first-party rate is unavailable.
Base entry
Radar
Full release historyKimi K2.6 release history
Radar confirmed these at the source. Use Kimi K2.6 in your work? Explore Radar to follow supported changes and choose your alerts.
Radar
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
- 256K
- 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
- Superseded
- API capabilities
- Tool calling, structured outputs, and batch support are not tracked yet
- Prompt caching
- Published at $0.16 per million cached input tokens
- Self-host
- Kimi K2.6 needs ~640GB VRAM (8× NVIDIA H100 (80GB)).
- Rate limits
- Not tracked yet
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 K2.6 ranks #53 of 218 on the public leaderboard with a score of 58.74/100. It does not yet have enough sourced coverage for a verified position.
Kimi K2.6 is a open weight model with a 256K context window. It uses an explicit reasoning mode, which can improve complex problem solving while adding latency and token use.
Official exact-value snapshot from MoonshotAI's April 20, 2026 Kimi K2.6 launch sources, supplemented by CursorBench v3.1. BenchLM maps directly comparable rows from the Hugging Face model card and Moonshot's Kimi tech blog. Unsupported internal showcase metrics remain excluded.
Its explicit predecessor is Kimi K2.5. 37 of 667 tracked benchmark slots currently have displayable evidence. Missing categories stay blank.
Its strongest eligible category is Mathematics at #1, while its lowest eligible position is Agentic at #57. particularly strong for mathematical reasoning, scientific computing, and quantitative analysis.
Last updated October 10, 2026. Runtime fields remain blank until a sourced snapshot exists.
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.
Self-host vs API cost
Estimates at 50,000 req/day · 1000 tokens/req average.
Kimi K2.6
API / mo$3,713
Self-host / mo$18,221
Break-even326M/day
Questions
How does Kimi K2.6 perform overall in AI benchmarks?
Kimi K2.6 ranks #53 out of 218 models on the public BenchAlign leaderboard, with a score of 58.74/100. Its evidence status is Estimated, and this profile shows 37 source-displayable benchmark rows. The label describes evidence depth, not a provider quality claim; inspect category rows before choosing a workload.
Is Kimi K2.6 good for knowledge and understanding?
Kimi K2.6 ranks #45 out of 177 eligible models for knowledge and understanding, with a public category score of 59.8/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.
Is Kimi K2.6 good for coding and programming?
Kimi K2.6 ranks #54 out of 146 eligible models for coding and programming, with a public category score of 43.9/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.
Is Kimi K2.6 good for mathematics?
Kimi K2.6 ranks #1 out of 7 eligible models for mathematics, with a public category score of 72.2/100. 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 K2.6 good for agentic tool use and computer tasks?
Kimi K2.6 ranks #57 out of 123 eligible models for agentic tool use and computer tasks, with a public category score of 43.7/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.
Is Kimi K2.6 good for multimodal and grounded tasks?
Kimi K2.6 ranks #27 out of 54 eligible models for multimodal and grounded tasks, with a public category score of 73.8/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.
Is Kimi K2.6 open source?
Kimi K2.6 is an open-weight model from Moonshot AI. Its weights can be downloaded for local or hosted deployment, subject to the published license. Open weight does not automatically mean open source: training data and training code may remain private, and commercial restrictions can still apply.
Does Kimi K2.6 have full benchmark coverage on BenchLM?
No. Kimi K2.6 currently has 55 source-displayable rows across 667 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 K2.6?
Kimi K2.6 has a reported context window of 256K 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.