When was Kimi K3 released?

Moonshot released Kimi K3 on 16 Jul 2026. It is available through the Opper gateway.

Does Kimi K3 support zero data retention?

Yes. Kimi K3 is available with zero data retention on Berget, DeepInfra, Fireworks, GreenPT, Modal, Novita, Sference, Tencent Cloud, TensorX, Wafer through the Opper gateway. Other routes carry zero data retention with abuse monitoring; see each route's window in the data-handling section.

Where is Kimi K3 hosted?

Kimi K3 runs on both EU and US hosting routes through the Opper gateway. Pick the route that matches your residency requirement.

Is Kimi K3 trained on prompts and outputs?

Training behaviour varies by route. See the per-route data-handling section for each provider's training posture.

Is a GDPR DPA available for Kimi K3?

Yes. A GDPR Data Processing Agreement is available on at least one route serving Kimi K3. Contact hello@opper.ai to receive it.

Kimi K3

by Moonshot

Released 34 days after Kimi K2.7 Code · All Moonshot releases

Kimi K3, released July 16, 2026, is Moonshot's largest model to date and a substantial step up in scale from the 1 trillion parameter Kimi K2 line. It uses a Mixture-of-Experts architecture with 2.8 trillion total parameters and 104 billion active per token, selecting 16 of 896 experts across 93 layers. The model targets long-horizon coding, knowledge work, and reasoning, scoring 93.5 on GPQA Diamond, 91.2 on BrowseComp, 88.3 on Terminal-Bench 2.1, and 67.5 on DeepSWE. Its 1,048,576 token context window is built for whole-repository work and extended agent runs, where the reasoning trace itself grows large. Vision is native, handled by the 401 million parameter MoonViT-V2 encoder, with 81.6 on MMMU-Pro. K3 runs in preserved thinking history mode, so multi-turn calls should pass complete assistant messages back including their reasoning content rather than stripping it. It is quantization-aware trained from the SFT stage onward using MXFP4 weights with MXFP8 activations, and its weights are published under the Kimi K3 License.

Call Kimi K3 on Opper with the OpenAI SDK. Sign up without a credit card. Get started

Get your API key Try in chat Compare

Key info

Input

Output

Features

Context window

Max output

Input price

$3.00 /1M$0.30 cache read

Output price

$15.00 /1M

Released

16 Jul 2026

EU residency available
US residency available
Zero data retention on pay-as-you-go
GDPR DPA available

Available routes

Kimi K3 runs on 17 different routes through the Opper gateway. Compare residency, zero data retention and training posture at a glance, with full data-handling detail per route below. In the EU it is served by Berget (Sweden), GreenPT, Sference and 1 more. See every Kimi model hosted in the EU. One route also sells priority processing: the same model at another price, picked per request with service_tier. How service tiers work.

Provider	Region	Training	Input	Output	Cache read
Arcee	US	No	$3.00	$15.00	$0.30	Not offered on this route
Berget	EU	No	$3.36	$16.81	Input price	Not offered on this route
DeepInfra	US	No	$2.85	$14.25	$0.28	Not offered on this route
DigitalOcean	US	No	$2.85	$14.25	$0.28	Not offered on this route
Fireworks	Multi	No	$3.00	$15.00	$0.30
GreenPT	EU	No	$3.70	$18.49	$0.92	Not offered on this route
Modal	Multi	No (opt-in)	$3.00	$15.00	$0.30	Not offered on this route
Moonshot AI	Multi	Yes	$3.00	$15.00	$0.30	Not offered on this route
Morph	US	No	$1.15	$12.83	$0.28	Not offered on this route
Morph	US	No	$6.00	$22.50	$0.60	Not offered on this route
Nebius	Multi	No	$3.00	$15.00	Input price	Not offered on this route
Novita	US	No	$3.00	$15.00	$0.30	Not offered on this route
Perplexity	Multi	—	$3.00	$15.00	$0.30	Not offered on this route
Sference	EU	No	$3.00	$15.00	$0.45	Not offered on this route
Tencent Cloud	Multi	No	$3.00	$15.00	$0.30	Not offered on this route
TensorX	EU	No	$3.00	$15.00	$0.75	Not offered on this route
Wafer	US	No	$3.00	$12.75	$0.30	Not offered on this route

Live routes and prices as JSON: api.opper.ai/v3/models?q=kimi-k3

Uptime and availability

Kimi K3 runs on 7 independently monitored routes through the Opper gateway. A direct integration leaves you on one of them, the gateway serves whichever is healthy.

100%effective uptime, last 30 days

Averaged across these routes on their own, a single provider reached 97.09% over the same window.

Daily status through the gateway, which serves each request from whichever route is healthy.

7 monitored providers serve Kimi K3.

Measured over the last 30 days from each provider's official status feed via StatusGator. Refreshed hourly. 9 further routes for this model are not on the board yet and are not counted above. See uptime for every provider Opper monitors.

Data handling per route

Each route hosting Kimi K3 has its own privacy posture, residency, and GDPR terms. Postures are maintained by Opper with a last-verification timestamp.

Arcee — United States🇺🇸

Zero data retention with abuse monitoring: prompts and outputs are kept for abuse monitoring only, for a window the provider does not state, and are not used for training. No training on customer data. US; SCCs; DPA available.

Zero data retention: With abuse monitoring, window not stated.
Training: No training on customer data.
Logging: Abuse monitoring
Abuse monitoring: On by default, holds flagged content
GDPR DPA: DPA available
Transfer mechanism: SCCs

Berget — Sweden🇸🇪

Zero data retention: nothing is logged, held for abuse monitoring or used for training. No training on customer data. EU; DPA available.

Zero data retention: Yes. Nothing is logged, held for abuse monitoring or used for training.
Training: No training on customer data.
Logging: None
Abuse monitoring: No classifier
Caching: Content cached for replay
Subprocessor access: No subprocessor reads content
GDPR DPA: DPA available
Transfer mechanism: Not applicable — data stays in EU

DeepInfra — United States🇺🇸

Zero data retention: nothing is logged, held for abuse monitoring or used for training. No training on customer data. US; unknown; DPA available.

Zero data retention: Yes. Nothing is logged, held for abuse monitoring or used for training.
Training: No training on customer data.
Logging: None
Abuse monitoring: No classifier
GDPR DPA: DPA available
Transfer mechanism: unknown

DigitalOcean — United States🇺🇸

Zero data retention with abuse monitoring: nothing is logged or used for training; content flagged by abuse monitoring is held for review. No training on customer data. US; dpf sccs; DPA available.

Zero data retention: With abuse monitoring, flagged content held.
Training: No training on customer data.
Logging: None
Abuse monitoring: Always on, holds flagged content, stored in United States
Caching: Content cached for replay
Subprocessor access: No subprocessor reads content
GDPR DPA: DPA available
Transfer mechanism: dpf sccs

Fireworks — Multi-region

Zero data retention: nothing is logged, held for abuse monitoring or used for training. No training on customer data. GLOBAL; SCCs; DPA available.

Zero data retention: Yes. Nothing is logged, held for abuse monitoring or used for training.
Training: No training on customer data.
Logging: None
Abuse monitoring: On by default, keeps nothing
Subprocessor access: Subprocessors may read content
GDPR DPA: DPA available
Transfer mechanism: SCCs

GreenPT — EU🇪🇺

Zero data retention: nothing is logged, held for abuse monitoring or used for training. No training on customer data. EU; DPA available.

Zero data retention: Yes. Nothing is logged, held for abuse monitoring or used for training.
Training: No training on customer data.
Logging: None
Abuse monitoring: Keeps nothing
Caching: Content cached for replay
Subprocessor access: Subprocessors may read content
GDPR DPA: DPA available
Transfer mechanism: Not applicable — data stays in EU

Moonshot AI — Multi-region

Content is retained: used for training by default. Trained on customer data by default. GLOBAL; unknown.

Zero data retention: Retained: used for training by default.
Training: Trained on customer data by default.
Logging: Abuse monitoring
Abuse monitoring: On by default, holds flagged content
Caching: Content cached for replay
GDPR DPA: No DPA
Transfer mechanism: unknown

Morph — United States🇺🇸

Zero data retention with abuse monitoring: prompts and outputs are kept for up to 30 days for abuse monitoring only, and are not used for training. No training on customer data. US; SCCs; DPA available.

Zero data retention: With abuse monitoring, 30-day window.
Training: No training on customer data.
Logging: Abuse monitoring (30-day retention)
Abuse monitoring: On by default, holds flagged content, a person can view samples
GDPR DPA: DPA available
Transfer mechanism: SCCs

Morph — United States🇺🇸

Zero data retention: With abuse monitoring, 30-day window.
Training: No training on customer data.
Logging: Abuse monitoring (30-day retention)
Abuse monitoring: On by default, holds flagged content, a person can view samples
GDPR DPA: DPA available
Transfer mechanism: SCCs

Nebius — Multi-region

Content is retained: provider logging. No training on customer data. GLOBAL; SCCs; DPA available.

Zero data retention: Retained: provider logging.
Training: No training on customer data.
Logging: Other
Abuse monitoring: No classifier
GDPR DPA: DPA available
Transfer mechanism: SCCs

Novita — United States🇺🇸

Zero data retention: nothing is logged, held for abuse monitoring or used for training. No training on customer data. US; SCCs; DPA available.

Zero data retention: Yes. Nothing is logged, held for abuse monitoring or used for training.
Training: No training on customer data.
Logging: None
Abuse monitoring: On by default, keeps nothing
GDPR DPA: DPA available
Transfer mechanism: SCCs

Perplexity — Multi-region

Content is retained: provider logging. Training posture pending review. GLOBAL; unknown.

Zero data retention: Retained: provider logging.
Logging: Other
Caching: Content cached for replay
Transfer mechanism: unknown

Sference — EU🇪🇺

Zero data retention: nothing is logged, held for abuse monitoring or used for training. No training on customer data. EEA; SCCs; DPA available.

Zero data retention: Yes. Nothing is logged, held for abuse monitoring or used for training.
Training: No training on customer data.
Logging: None
Abuse monitoring: No classifier
Caching: Content cached for replay
Subprocessor access: No subprocessor reads content
GDPR DPA: DPA available
Transfer mechanism: SCCs

Tencent Cloud — Multi-region

Zero data retention: nothing is logged, held for abuse monitoring or used for training. No training on customer data. GLOBAL; SCCs; DPA available.

Zero data retention: Yes. Nothing is logged, held for abuse monitoring or used for training.
Training: No training on customer data.
Logging: None
Abuse monitoring: On by default, keeps nothing
Caching: Content cached for replay
GDPR DPA: DPA available
Transfer mechanism: SCCs

TensorX — European Union🇪🇺

Zero data retention: nothing is logged, held for abuse monitoring or used for training. No training on customer data. EU; DPA available.

Zero data retention: Yes. Nothing is logged, held for abuse monitoring or used for training.
Training: No training on customer data.
Logging: None
Abuse monitoring: No classifier
Caching: Content cached for replay
GDPR DPA: DPA available
Transfer mechanism: Not applicable — data stays in EU

Wafer — United States🇺🇸

Zero data retention: nothing is logged, held for abuse monitoring or used for training. No training on customer data. US; SCCs; DPA available.

Zero data retention: Yes. Nothing is logged, held for abuse monitoring or used for training.
Training: No training on customer data.
Logging: None
Abuse monitoring: No classifier
GDPR DPA: DPA available
Transfer mechanism: SCCs

Benchmarks

Independent benchmark scores — composite indices for reasoning, coding, and math, plus individual eval scores where available.

Global rank#42 of 687 LLMs

TierStrong

Output speed40 tok/s

First token2.57s

Intelligence Index43.6

Coding Index76.2

Reasoning & knowledge

GPQA Diamond

94%

Humanity's Last Exam

47%

Long-context reasoning

89%

Coding

SciCode

60%

See full leaderboard →Benchmarks via Artificial Analysis · View on AA

Get started

Call Kimi K3 through the Opper gateway with one API key. Opper is drop-in compatible with the OpenAI, Anthropic and Google AI SDKs.

Kimi K3 is a premium model on Opper. Sign up needs no credit card: you get an API key straight away and the free models work in the playground and the API. Add a card to use premium models, pay-as-you-go with no minimum.

Set it up with your agent

Copy this and paste it into a coding agent like Claude Code, Cursor or Codex and it'll wire up Opper for you.

Or call it directly

1import OpenAI from "openai";
2
3const client = new OpenAI({
4  apiKey: process.env.OPPER_API_KEY,
5  baseURL: "https://api.opper.ai/v3/compat",
6});
7
8const completion = await client.chat.completions.create({
9  model: "kimi-k3",
10  messages: [{ role: "user", content: "Hello" }],
11});
12console.log(completion.choices[0].message.content);