Is Kimi K3 available with Zero Data Retention on Elosia?

Kimi K3 runs on Elosia's standard backend; ZDR is available on many other models in the catalog.

How much does Kimi K3 cost on Elosia?

Kimi K3 is billed by usage at $3 per 1M input tokens and $15 per 1M output tokens, within your Elosia plan.

What is Kimi K3's context window?

Kimi K3 supports a context window of up to 1,048,576 tokens.

Is Kimi K3 hosted in the EU?

Kimi K3 is not available on Elosia's EU-hosted (EdenAI) backend; it runs on the default backend.

When should you choose Kimi K3?

Kimi K3 is a flagship model, best for the most demanding reasoning, agentic and long-context tasks.

Kimi K3

MoonshotAiFlagship

ThinkingTool UseVisionStructured Output

About this model

Kimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI. It is suited for complex coding, knowledge work, and long-horizon agentic workflows, and is particularly strong at...

Performance Tier

Flagship

Kimi K3 is a flagship model from MoonshotAi : the most capable in their lineup.

Best-in-class model from this provider. Highest performance across benchmarks, ideal for demanding tasks.

Pricing

This model is included in Elosia plans

High

High pricing. Reserve this level for tasks that demand maximum quality.

Type	per 1M tokens
Input (prompt)	$3.00
Output (completion)	$15.00
Cache read	$0.300

Capabilities

Context Length1.0M

Max Output Tokens—

TokenizerOther

Inputtext, image

Outputtext

Release DateJuly 16, 2026

Benchmarks

General Intelligence

MMLU

Not reported

GPQA Diamond

93.5%

Mathematics

MATH-500

Not reported

Programming

HumanEval

Not reported

SWE-bench Verified

Not reported

Reasoning

IFEval

Not reported

Humanity's Last Exam

56%

Multimodal

MMMU-Pro

81.6%

Agentic

Terminal-Bench 2.1

88.3%

Where does Kimi K3 stand?

Compare its performance index against every other model.

View the performance leaderboard

Recommended Use Cases

CodingAnalysisResearchGeneral Chat

Strengths

Moonshot flagship at 2.8T parameters, activating 16 of 896 experts per token (LatentMoE)
Graduate-level reasoning at the top of the field (GPQA Diamond 93.5%), independently confirmed by Artificial Analysis
1M-token context, four times the 256K window of the Kimi K2 family
Native vision capabilities, with strong multimodal reasoning (MMMU-Pro 81.6%)
Kimi Delta Attention with attention residuals, for roughly 2.5x the scaling efficiency of Kimi K2 per Moonshot

Limitations

Moonshot official evaluation focused on agentic, reasoning and vision benchmarks: MMLU, MATH-500 and HumanEval are not part of the reported suite
Thinking is always on and reasoning_effort accepts only "max", so there is no instant path for simple calls
Most expensive model in the Kimi line by a wide margin ($3/M input, $15/M output)
Terminal-Bench 2.1 is self-reported at 88.3%, while independent runs land lower (Artificial Analysis 85.0%, Vals AI 80.9%)