Text model

Kimi by Moonshot AI — a million tokens of context

Q: Which Kimi models are available in GPTunneL?

The current list is always visible in the «Every Kimi model» block above — it is built straight from the platform catalogue.

Q: What is a million-token context actually for?

It is roughly fifteen hundred pages of text in a single request: a contract with every annex, a year of project correspondence or a large repository in full. The model sees all of it at once and does not lose the beginning by the time it reaches the end.

Q: How is K2.7 Code different from the regular model?

It is the same family fine-tuned on code: it edits files more carefully, holds the structure of a project better and is more confident in large repositories.

Q: Which languages does Kimi handle?

The models work comfortably in dozens of languages — they understand the request and answer in the same language.

Q: Does my data go to the model's developer?

No. Requests go through GPTunneL's infrastructure and the model runs on its capacity, not on the developer's servers.

Kimi models by Moonshot AI in GPTunneL: a 1M token context, reasoning and open weights. Pay per token, no subscription and no VPN.

Try Kimi Available models

What the Kimi family is good at

The traits Moonshot AI keeps from generation to generation — you can count on them whichever model you pick.

A very long context

The main reason people pick this family: the top model holds a million tokens — a folder of documents, a year of correspondence or an entire repository, with no slicing into chunks.

Open weights

Moonshot publishes the weights of its flagships under a free licence, so they can be downloaded and run on your own hardware. In GPTunneL the same models are available without any hardware of your own.

Reasoning mode

The model spends more time on a hard task and reaches the answer via intermediate steps — it helps noticeably with maths, logic and reading code.

Works with tools

The family can search the web, call tools and follow a multi-step plan — which is why it gets picked for agentic work.

Which model to pick

Three current models in the family — for different tasks and budgets.

	Kimi K2.6Everyday tasks	Kimi K3The default pick	Kimi K2.7 CodeDevelopment
Best for	Text, correspondence, reading documents	Hard reasoning, large document sets, agentic runs	Edits inside a project, reading unfamiliar code, review
Context	262K tokens	1M tokens	262K tokens
Reasoning	Yes	Yes	Yes
Price per token	Low	Above average	Low

Specs come from the platform catalogue. Exact per-model prices are on the Pricing

Every Kimi model

Release dates follow Moonshot AI's official announcements.

July 2026

Kimi K3

The current generation and the largest open-weight model on the market: 2.8 trillion parameters, a million-token context and image understanding. Moonshot published the weights a week and a half after launch.

2026

Kimi K2.7 Code Kimi K2.6 Kimi K2.5

Interim updates to the second generation: context grew to 262 thousand tokens, reasoning and image understanding arrived, and a separate Code branch was split off for development work.

November 2025

Kimi K2 Thinking

The K2 variant with long reasoning — it thinks a task through for longer and shows its train of thought. This is where the reasoning mode entered the family.

July 2025

Kimi K2

The generation that put Moonshot on the map outside China: an open trillion-parameter model on par with closed flagships.

How billing works

There is no subscription: you top up one balance and spend it on any model on the platform.

Pay per token

You are charged for exactly what the request and the answer used. Not using it costs nothing: no limits, no monthly fee.

One balance for every model

Kimi, ChatGPT, Claude, image and video generation — all out of the same wallet. No separate subscription per service.

Top up the way you prefer

International cards, Apple Pay, Google Pay or crypto. The minimum top-up is $5.

See pricing

How to start with Kimi

One account for every model on the platform.

Emailyou@example.comPassword••••••••

Build a table of discrepancies by quarter from these reports and point out where the numbers do not add up

q1-q4.pdfKimi K3

Discrepancies by quarter

In Q2 the summary revenue is 4.2M higher than the sum across regions — the gap sits entirely in the «other» line.

Q3 reconciles except for the FX revaluation, which was taken at the period-end rate rather than the deal date.

Q4 counts the December refund twice, and that is what inflates the full-year total.

The model switches right inside the input — you can change it mid-conversation.

The answer arrives in the chat and the context is kept until the conversation ends.

Try Kimi in GPTunneL

Signing up takes a minute, and $5 on the balance is enough to see whether the model fits your task.

Try Kimi

Frequently asked questions

/01Do I need a VPN to use Kimi through GPTunneL?

No. Requests go through GPTunneL's infrastructure, so the models open like any ordinary website — no VPN and no proxy.

/02Which Kimi models are available in GPTunneL?

/03What is a million-token context actually for?

/04How is K2.7 Code different from the regular model?

/05Which languages does Kimi handle?

/06Does my data go to the model's developer?

Kimi is a family of neural networks from the Chinese company Moonshot AI, known above all for two things: a very long context and open weights. The K3 flagship holds a million tokens in context and is published under a free licence — at launch it was the largest open-weight model in the world.

The family answers questions, writes and edits text, works with code and reads images. The larger models can reason — think a hard task through and reach the answer via intermediate steps — and can also search the web and call tools, which is why they get picked for multi-step agentic work. For development there is a separate Code branch.

Kimi models are available in GPTunneL — worldwide, without a VPN, with one balance and billing for the tokens you actually spend.