VENICE ATLAS · MODEL DOSSIER · KIMI K3

Model overview

Kimi K3

Name: Kimi K3
Brand: venice.ai
Price: 3.7500 USD
Availability: InStock

Kimi K3 is an ultra-large-scale, 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 navigating large repositories, using tools, debugging, and iterating against images, logs, tests, and runtime feedback.

kimi-k3 · venice.ai

Text · Venice catalog$3.75 per 1M tokensPrice on Venice AI

ModalityTextOn Venice AI

Primary price$3.75 per 1M tokensper 1M tokens

Context window1MMax tokens

Benchmarks31Benchmark results

Benchmark results

See the tests behind each score. They use the benchmark’s original scoring scale.

Agents#1237.25 rating Biology#737.92 rating Chemistry#937.92 rating Coding#839.85 rating Vision#936.03 rating General intelligence#1051.83 rating Math#941.25 rating Multimodal#537.21 rating Physics#937.92 rating Productivity#108.53 rating Reasoning#851.14 rating Search#134.66 rating Systems#710.66 rating Tool Calling#630.95 rating

Agents11 results

AA-Briefcaseaa-briefcase · agents

51.6%

ReportedOct 2026

AutomationBenchautomationbench · agents

30.8%

ReportedOct 2026

DECK-Benchdeck-bench · agents

73.5%

ReportedOct 2026

Job Benchjob-bench · agents

52.9%

ReportedOct 2026

Kimi Code Bench v2kimi-code-bench-v2 · agents

72.9%

ReportedOct 2026

MLS-Bench Litemls-bench-lite · agents

48.3%

ReportedOct 2026

OfficeQA Proofficeqa-pro · agents

63.3%

ReportedOct 2026

PostTrainBenchposttrainbench · agents

36.6%

ReportedOct 2026

Program Benchprogram-bench · agents

77.8%

ReportedOct 2026

SpreadsheetBench 2spreadsheetbench-2 · agents

34.8%

ReportedOct 2026

Toolathlontoolathlon · agents

73.2%

ReportedOct 2026

Coding4 results

DeepSWEdeepswe · code

67.5%

ReportedOct 2026

DeepSWE 1.1deepswe-1.1 · code

69.0%

ReportedOct 2026

FrontierSWEfrontierswe · code

81.2%

ReportedOct 2026

SWE-Marathonswe-marathon · code

42.0%

ReportedOct 2026

Math1 result

MathVisionmathvision · math

97.8%

ReportedOct 2026

Reasoning7 results

APEX-Agentsapex-agents · reasoning

37.6%

ReportedOct 2026

BrowseCompbrowsecomp · reasoning

91.2%

ReportedOct 2026

CharXiv-Rcharxiv-r · reasoning

91.3%

ReportedOct 2026

GPQAgpqa · reasoning

93.5%

ReportedOct 2026

Humanity's Last Examhumanity's-last-exam · reasoning

56.0%

ReportedOct 2026

MCP Atlasmcp-atlas · reasoning

84.2%

ReportedOct 2026

Terminal-Bench 2.1terminal-bench-2.1 · reasoning

88.3%

ReportedOct 2026

Search1 result

DeepSearchQAdeepsearchqa · search

95.0%

ReportedOct 2026

Vision6 results

BabyVisionbabyvision · vision

85.7%

ReportedOct 2026

MMMU-Prommmu-pro · vision

81.6%

ReportedOct 2026

MMMU-Pro (with tools)mmmu-pro-with-tools · vision

83.4%

ReportedOct 2026

OmniDocBenchomnidocbench · vision

91.1%

ReportedOct 2026

PerceptionBenchperceptionbench · vision

58.5%

ReportedOct 2026

WorldVQAworldvqa · vision

51.0%

ReportedOct 2026

Learn what each benchmark measures ↗

Pricing detail

Input price per 1M tokens: $3.75
Output price per 1M tokens: $18.75
Cache read price per 1M tokens: $0.38
Blended price per 1M tokens (50/50): $11.25

Prices are published by Venice AI. Venice model page ↗

Capabilities

ReasoningFunction callingVisionWeb search

Endpoint metadata

Privacy: private
Quantization: not-available
Maximum completion tokens: 131K
Uncensored: No
Beta: No
Mapping confidence: 100.0%

View model source ↗

All capabilities

Max Images: 10
Quantization: not-available
Supports E2 E E: No
Supports Vision: Yes
Supports X Search: No
Optimized For Code: Yes
Supports Log Probs: No
Supports Reasoning: Yes
Supports Web Search: Yes
Supports Audio Input: No
Supports Video Input: No
Default Reasoning Effort: max
Reasoning Effort Options: none, low, high, max
Supports Multiple Images: Yes
Supports Response Schema: Yes
Supports Tee Attestation: No
Supports Function Calling: Yes
Supports Reasoning Effort: Yes

Traits

default_reasoning

Related models

More from moonshotai

kimi-k2-5Kimi K2.5$0.56 input per 1M tokens ↗kimi-k2-6Kimi K2.6$0.75 input per 1M tokens ↗kimi-k2-7-codeKimi K2.7 Code$0.75 input per 1M tokens ↗