Is Kimi K2 Thinking open source?

It is MoonshotAI's most advanced open reasoning model to date, released in the K2 series. That means the weights can be run on your own infrastructure or through third-party inference providers, as well as via MoonshotAI's own API.

When should I use Kimi K2 Thinking instead of a standard K2 model?

Use it when the task has multiple dependent steps, such as agentic research, tool-calling chains or long code work, where an error early on spoils everything after it. For short, high-volume tasks the extra reasoning tokens are cost without benefit.

Can Kimi K2 Thinking handle images?

No. It is a text-only model, so any workflow involving screenshots, diagrams or document images needs a multimodal model instead.

Does Kimi K2 Thinking support long documents?

Yes. Its context window holds large document sets, full codebases or extended agent traces in a single pass, and its output ceiling is high enough that long reasoning plus a long final answer does not get truncated.

Moonshot AI model°

Kimi K2 Thinking

Kimi K2 Thinking is MoonshotAI's open reasoning model, the step in the K2 series built for agentic, long-horizon work rather than single-turn answers.

Moonshot AIVerified 24/09/2026Released 06/11/2025

01 / decision

Decision snapshot

Each figure sits next to the middle of the field, so it can be read as dear or cheap, wide or narrow, rather than floating on its own. Compared against the 126 models we track, not against an absolute standard.

Capability

Not measured

No published benchmark scores yet

Input / 1M tokens

$0.60

field median $0.30

Output / 1M tokens

$2.50

field median $1.25

Context

262K

field median 524K

02 / overview

What Kimi K2 Thinking is, and when to reach for it

Four questions, answered separately, because somebody arrives at one of them rather than at the top of the page.

What it is

The job it was built for, and the job it is not for.

Kimi K2 Thinking is a text-only reasoning model aimed at tasks that run over many steps: multi-stage research, tool-calling chains, long refactors, workflows where the model has to hold a plan together rather than reply once. It is the reasoning branch of the K2 family, so it spends tokens thinking before it answers. It is not a multimodal model and not the right choice for short, latency-sensitive chat or anything involving images.

When it arrived, and when to use it

Where it sits in its line, and when a sibling is the better pick.

Kimi K2 Thinking first appeared in November 2025, extending MoonshotAI's K2 series from general-purpose into explicit long-horizon reasoning. Pick it over the non-thinking K2 models when the task has several dependent stages and a wrong early step ruins the output. For high-volume classification, extraction or routing, the reasoning overhead is wasted and a plainer model is the better call.

How you reach it

The API, the apps it powers, and what its limits let you do.

Kimi K2 Thinking is reached through MoonshotAI's API and through the open-weight distributions, so it can be self-hosted or run via third-party inference providers. Its context window comfortably holds a large codebase, a full document set or a long agent trace, and the unusually generous output ceiling means it can produce extended reasoning plus a long final answer without being cut off mid-task. Text in, text out, with tool calling for agent loops.

Why it matters

What changes because this exists, or why it does not.

An open reasoning model with this much room for both input and output makes long agentic runs practical to build and inspect without committing to a closed API. Teams that previously chunked long jobs into stitched-together calls, or watched a run truncate before it finished, can hand the whole thing over in one pass. Being open weights also means the reasoning trace and deployment sit inside your own infrastructure if that matters.

03 / evidence

How much of this is verified

Split by category, so a strong number never hides a thin evidence base. Verified means we read it on the benchmark's own published results; a provider's claim about its own model is shown and labelled rather than dropped.

No published benchmark scores for this model yet.

We publish a score only where we can link the result to where it was published. Until a benchmark result for this model exists in a source we read, this section stays empty rather than being filled with a provider's marketing figure.

How we decide what counts as evidence

04 / ledger

Benchmark ledger

Every published row, grouped by category, each compared with the best published score on the same benchmark. 'Is 64% good' is a question nobody can answer; '26 points behind the leader' is one anybody can.

Nothing in the ledger yet.

Each row here carries a score, the benchmark it came from, what the leading model scored on the same test, and a link to the published result. Rows appear as results are published and read.

How we decide what counts as evidence

05 / capability

Capability shape

Where this model is strong, and against how many peers. Ranks are against models with evidence in that category, not against everything we track: ranking against models nobody tested would rank who published, not who is better.

No category scores to shape yet.

A category score is the weighted mean of the benchmarks published for it. With no published rows there is nothing to average, and an empty chart drawn at zero would say something false.

How we decide what counts as evidence

06 / cost

What it costs

List API rates as last read from the provider, with the source on every row, plus every change we have recorded since we started tracking it.

Kimi K2 Thinking API pricingSurge45°
Charge	Price	Unit	Read on	Source
Input	$0.60	per 1M tokens	2026-09-24	Check
Output	$2.50	per 1M tokens	2026-09-24	Check

Kimi K2 Thinking is priced at the cheap end of the reasoning tier, closer to a general-purpose model than to the frontier closed reasoners it is aimed at. That matters more than usual here, because reasoning models burn output tokens by design, and a model that thinks at length on expensive tokens gets costly fast. Self-hosting the open weights changes the economics again for teams with the hardware.

What a month costsSurge45°
Workload	Input / month	Output / month	Cost
A small product team	20M tokens	5M tokens	$24.50
A busy support assistant	200M tokens	40M tokens	$220.00
A document pipeline	1000M tokens	100M tokens	$850.00

List API rates, no caching and no batch discount, which both providers offer and which change the answer a great deal. Treat these as the ceiling, not the bill.

07 / specs

Specifications

As listed by the provider's own catalogue and re-read every few hours. Anything absent is absent there too.

SpecificationSurge45°
Context window	262,144 tokens
Maximum output	98,304 tokens
Modalities	text
Released	06/11/2025
Status	Current
Catalogue identifier	moonshotai/kimi-k2-thinking

08 / lineage

Lineage

What this model replaced, what replaced it, and what else its provider has in the field.

Also from Moonshot AI

Kimi K2 090504/09/2025

09 / line

The line

Every model in this family in release order, so a page from eight months ago says in one glance that two newer ones exist.

Nothing else in this line yet.

A line is worked out from the naming and the release dates across every model page we hold. It fills in as the provider ships successors, or as we pick up the models that came before this one.

How we decide what counts as evidence

10 / notes

Our notes

What this model changes for a brand trying to be cited in AI answers, and every change we have logged since it launched.

What it changes for you

Reasoning models with long context read more of a source before answering, so a thin comparison page or a claim without supporting detail is likelier to be skipped in favour of a page that survives scrutiny. When an agent is running a multi-step research task about your category, it will visit several sources and reconcile them, which rewards brands whose pricing, integrations and limits are stated consistently across their own site. Open weights also mean this model turns up inside products you have never heard of, so citation visibility is no longer just a question of the big three assistants.

Where buyers meet this model

Most buyers meet Kimi K2 Thinking through the Kimi consumer app and web assistant, where it powers the deeper research and reasoning modes. Developers meet it through MoonshotAI's API or through the open weights on third-party inference platforms, which is where it shows up inside other products. It is increasingly the model sitting behind agentic research tools built by teams who want open weights rather than a closed provider.

Change log

Nothing published here yet. Changes appear within hours of a provider announcing them.

11 / questions

Questions

The things people ask about this model, answered from what is on this page rather than from anywhere else.

Is Kimi K2 Thinking open source?: It is MoonshotAI's most advanced open reasoning model to date, released in the K2 series. That means the weights can be run on your own infrastructure or through third-party inference providers, as well as via MoonshotAI's own API.
When should I use Kimi K2 Thinking instead of a standard K2 model?: Use it when the task has multiple dependent steps, such as agentic research, tool-calling chains or long code work, where an error early on spoils everything after it. For short, high-volume tasks the extra reasoning tokens are cost without benefit.
Can Kimi K2 Thinking handle images?: No. It is a text-only model, so any workflow involving screenshots, diagrams or document images needs a multimodal model instead.
Does Kimi K2 Thinking support long documents?: Yes. Its context window holds large document sets, full codebases or extended agent traces in a single pass, and its output ceiling is high enough that long reasoning plus a long final answer does not get truncated.

Surge45°

Is Kimi K2 Thinking recommending you?

Models change what gets cited. We measure whether AI answers name your brand or your competitors across every assistant, and show you what to change.

Free AI visibility audit Book a call