Kimi K2.6
Moonshot AI · released Apr 14, 2026 · moonshotai/Kimi-K2.6
- Type
- Open weightsCustom licence
- Params
- 1.1T
- Context
- 262K
32B active per word · about 197K words of context · download allowed, licence restricts use
Our take
Written Sep 2, 2026Kimi K2.6 is a large mixture-of-experts model from Moonshot AI with 32 billion active parameters per word and a 262,144-token request limit. It excels at mathematics and coding on refreshed leaderboards, but its agentic scores sit below neutral and its licence carries restrictions.
Pick this for demanding mathematics or standard coding workloads where LiveBench scores matter, or for long-document analysis at a quarter-million tokens. Use it if you need high throughput and can route to the fastest host. Skip it if you need autonomous agent reliability, unrestricted redistribution, or consistent speed across providers.
The case for it
- Elite mathematics on refreshed competition tasks: 84.28% on LiveBench.
- Strong coding across rubric and human-preference boards: 78.57% LiveBench and 1514.4 Arena Coding.
- 262,144-token request limit, far above most mid-size models.
- Up to 110.5 tokens per second on the fastest tracked host.
The case against it
- Agentic recovery and task outcome below neutral on Arena Agent; agentic coding lags standard coding by 31.65 points.
- Custom restricted licence, not Apache or MIT.
- Throughput varies 4.6× across providers, from 24 to 110.5 tokens per second.
How good is it?
An open-weights text model for everyday questions, coding and tool use, though it can struggle to recover after a failed step.
- answering everyday questionsArena Text (overall) · 37th of 168
- writing and completing codeArena Coding · 29th of 168
- calling tools to carry out requestsArena Agent · Tool use · 2nd of 55
- changing course when given new instructionsArena Agent · Steerability · 5th of 55
- getting back on track after a step failsArena Agent · Recovery · 46th of 55
EverydayGeneral questions and everyday reasoning
Arena Text (overall)37th of 168 · 1461
CodingWriting and fixing code on its own
Arena Coding29th of 168 · 1516
AgenticPlanning, calling tools, staying on task
Arena Agent28th of 55 · −0.003
WritingDrafting and rewriting prose
Arena Creative Writing44th of 168 · 1433
Placings on Arena's agent boards, from live sessions people ran themselves. A model can lead on one of these and sit mid-field on the others.
Boards this model appears on that none of the ratings above are built on.
Every published score for this model20 scoresEvery figure we hold, from 20 boards, with who ran it and a link to the source — including the boards no rating above is built on.
Can you run it yourself?
GeForce RTX 4090 · 24 GB
Too large for this card. The weights do not fit even with part of them offloaded to system memory.
Apple M1 Pro (16-core GPU) · 32 GB
Too large for this card. The weights do not fit even with part of them offloaded to system memory.
Radeon RX 7900 XT · 20 GB
Too large for this card. The weights do not fit even with part of them offloaded to system memory.
Memory use by level
Against a 24 GB card.
What is quantisation? →This model on every device we track71 devicesThe Q4 build most people download, on each device: what the weights come to, how much context the memory leaves, and whether it runs. Smallest device that runs it first.
Check against your own machine → · Where to rent it hosted →
Or rent it from someone else
Prices checked 1 hour ago — each listing carries its own date.
- per 1M tokens
- $0.43 in / $1.83 out
- Context served
- 262K
- Throughput
- Not measured
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| OpenRouterOpenRouter's own listing | $0.43 / $1.83checked 1 hour ago | 262K | not measured | Unknown | Unknown | Unknown |
| Baidufp4Through OpenRouter | $0.43 / $1.83checked 1 hour ago | 262K236K max reply | 51 tok/s | No | Yesunknown period | Unknown |
| DigitalOcean GradientThrough OpenRouter | $0.57 / $2.40checked 1 hour ago | 262K236K max reply | 36 tok/s | No | No | Confirmed |
| Inceptronint4Through OpenRouter | $0.43 / $2.45checked 1 hour ago | 262K236K max reply | 46 tok/s | No | No | Confirmed |
| Decartfp4Through OpenRouter | $0.59 / $2.47checked 1 hour ago | 262K236K max reply | 55 tok/s | No | No | Confirmed |
| StreamLakefp8Through OpenRouter | $0.60 / $2.52checked 1 hour ago | 256K230K max reply | 62 tok/s | No | Yesunknown period | Unknown |
| Chutesint4Through OpenRouter | $0.50 / $2.85checked 1 hour ago | 262K66K max reply | 47 tok/s | No | Yesunknown period | Unknown |
| SiliconFlowfp8Through OpenRouter | $0.77 / $3.40checked 1 hour ago | 262K236K max reply | 24 tok/s | No | No | Confirmed |
| Novita AIDirect and through OpenRouter | $0.80 / $3.40checked 1 hour ago | 262K236K max reply through OpenRouter | 31 tok/sthrough OpenRouter | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterConfirmed |
| CoreWeavefp4Through OpenRouter | $0.65 / $3.41checked 1 hour ago | 262K236K max reply | 102 tok/s | No | No | Confirmed |
| DeepInfrafp4Direct and through OpenRouter | $0.75 / $3.50checked 1 hour ago | 262K16K max reply through OpenRouter | 21 tok/sthrough OpenRouter | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterConfirmed |
| Venice AIint4Through OpenRouter | $0.75 / $3.50checked 1 hour ago | 256K66K max reply | 22 tok/s | No | No | Confirmed |
| Crusoebf16Through OpenRouter | $0.70 / $3.50checked 1 hour ago | 262K236K max reply | 77 tok/s | No | No | Confirmed |
| Parasailint4Through OpenRouter | $0.75 / $3.50checked 1 hour ago | 262K236K max reply | 66 tok/s | No | No | Confirmed |
| GMICloudfp8Through OpenRouter | $0.85 / $3.60checked 1 hour ago | 262K236K max reply | 21 tok/s | No | Yesunknown period | Unknown |
| AtlasCloudint4Through OpenRouter | $0.95 / $4.00checked 1 hour ago | 262K236K max reply | 95 tok/s | No | Yesunknown period | Unknown |
| Moonshot AIint4Through OpenRouter | $0.95 / $4.00checked 1 hour ago | 262K236K max reply | 46 tok/s | No | No | Confirmed |
| Cloudflare Workers AIThrough OpenRouter | $0.95 / $4.00checked 1 hour ago | 262K236K max reply | 35 tok/s | No | Yesunknown period | Unknown |
| PhalaThrough OpenRouter | $1.09 / $4.60checked 1 hour ago | 262K236K max reply | 27 tok/s | No | No | Confirmed |
Baidu
DigitalOcean Gradient
Novita AI
DeepInfra
Cloudflare Workers AIAcross the 19 listings we hold: 18 say they do not train on prompts (2 of them only through OpenRouter), 0 say they do and 1 does not say. 12 appear in the zero-retention registry we check (2 of them only through OpenRouter); the rest are unknown to us.
What each host's API supports
From the parameter list each endpoint publishes. Streaming is omitted: nothing we hold reports it, for any model.
| Provider | Tool calling | JSON output | Strict schema |
|---|---|---|---|
| OpenRouterOpenRouter's own listing | ✓ | ✓ | ✓ |
| Baidufp4Through OpenRouter | ✓ | ✓ | ✓ |
| DigitalOcean GradientThrough OpenRouter | ✓ | ✓ | ✓ |
| Inceptronint4Through OpenRouter | ✓ | ✓ | ✓ |
| Decartfp4Through OpenRouter | ✓ | ✓ | ✓ |
| StreamLakefp8Through OpenRouter | ✓ | ✓ | ✓ |
| Chutesint4Through OpenRouter | ✓ | ✗ | ✓ |
| SiliconFlowfp8Through OpenRouter | ✓ | ✓ | ✓ |
| Novita AIDirect and through OpenRouter | ✓ | ✓ | ✓ |
| CoreWeavefp4Through OpenRouter | ✓ | ✓ | ✓ |
| DeepInfrafp4Direct and through OpenRouter | ✓ | ✓ | ✓ |
| Venice AIint4Through OpenRouter | ✓ | ✓ | ✓ |
| Crusoebf16Through OpenRouter | ✓ | ✓ | ✓ |
| Parasailint4Through OpenRouter | ✓ | ✓ | ✓ |
| GMICloudfp8Through OpenRouter | ✓ | ✓ | ✓ |
| AtlasCloudint4Through OpenRouter | ✓ | ✓ | ✓ |
| Moonshot AIint4Through OpenRouter | ✓ | ✓ | ✓ |
| Cloudflare Workers AIThrough OpenRouter | ✓ | ✓ | ✓ |
| PhalaThrough OpenRouter | ✓ | ✓ | ✓ |
Tool calling: 19 of 19 listings say yes. JSON output: 18 of 19 listings say yes, 1 says no. Strict schema: 19 of 19 listings say yes.
Models people weigh against Kimi K2.6
When we formed this view
Recent changes
What moved
input −3% ($0.453 → $0.438 per 1M tokens), cache read −2% ($0.123 → $0.121 per 1M tokens)What moved
input +11% ($0.409 → $0.455 per 1M tokens), output +3% ($2.39 → $2.45 per 1M tokens), cache read +39% ($0.089 → $0.124 per 1M tokens)What moved
input −6% ($0.434 → $0.409 per 1M tokens), cache read −15% ($0.105 → $0.089 per 1M tokens)What moved
input −4% ($0.45 → $0.43 per 1M tokens), output −20% ($2.97 → $2.39 per 1M tokens), cache read −17% ($0.126 → $0.105 per 1M tokens)Each date is the day we first saw the change, or the day the maker announced it.
What we do not know about this model yet
- We hold no measured file for it, so all 3 sizes on this page are calculated from the parameter count.
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- 1 of 19 listings does not say whether it trains on prompts, and 2 answer only through OpenRouter, not for their own listing.
- We hold no batch or off-peak rate for any of its listings.
- We hold a decode speed for it, but no prompt-processing (prefill) figure, so how long the input side of a job takes is unknown to us.
Licence and identifiers
What the licence allowsCustom licence, what it allows commercially, and the identifiers you need to pull this model — its Hugging Face repo, our slug and a machine-readable card.
Licence
Custom licence
This model ships custom license terms that don't map to a known template. We haven't parsed them, so commercial use, redistribution and derivatives are unverified — review the original terms before shipping.
Identifiers
- Hugging Face
- moonshotai/Kimi-K2.6
- Architecture
- Mixture of experts
- Takes in, gives back
- Text and images in, text out
- Catalogue slug
- moonshotai-kimi-k2-6