Skip to main content

Kimi K3

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 navigating large repositories, using tools, debugging, and iterating against images, logs, tests, and runtime feedback. Its architecture uses KDA and Attention Residuals for computational efficiency.

Model card on Hugging Face
Input / output price
$1.85 / $9.50 per 1M
Context
1M tokens
Released
Jul 16, 2026
Agent setup

Paste into Claude Code, Cursor, or Codex

Help me set up the Respan gateway: fetch and follow https://www.respan.ai/docs/agent-setup.md. Use model "moonshot/kimi-k3".

Read docs

Providers

23 routes from 22 providers, cheapest first, OpenRouter last, priced per 1M tokens after gateway discounts.

ProviderInputOutput
Concentrateconcentrate/kimi-k3
$0.60/M$15.00/M
Waferwafer/kimi-k3
$0.60/M$15.00/M
Makoramakora/kimi-k3
$1.00/M$18.00/M
Chuteschutes/kimi-k3
$1.85/M$9.50/M
Relacerelace/kimi-k3
$2.00/M$14.00/M

Is a provider down? AWS Bedrock status, Fireworks AI status, Kimi status, Perplexity status, OpenRouter status

Pricing

Averaged across 23 routes from 22 providers, in USD per 1M tokens after gateway discounts.

Average input
$2.46 per 1M
Average output
$14.95 per 1M
Average cache read
$0.45 per 1M
Average cache write
$2.91 per 1M

Cost by volume

Total cost from 0 to 1T tokens on the cheapest and most expensive provider.

Highest$3.30/MBedrockLowest$0.192/MOpenRouter

Throughput and latency

30-day averages from requests through the Respan gateway. Under each figure, every model on the gateway by its average, this one in blue.

Output speed

Higher is better

29tokens/sacross providers

Faster than 56% of the other 596 models on the gateway

Time to first token

Lower is better

1.8sacross providers

Faster than 47% of the other 153 models on the gateway

Response latency

Lower is better

15sacross providers

Faster than 14% of the other 776 models on the gateway

Benchmarks

The Respan Index plus 82 published scores from 14 sources, each linked to where it was published.

52.8/ 100

Respan Index#35 of 176

176 ranked models. The shaded band is this model's 95% interval, 47.9 to 57.8. Grok 4.7 and Qwen3.8-Max fall inside it, so the gap is within the margin of error.
Kimi K3 by category: Coding 38.8, Agents 61.8, Reasoning 54.8, Knowledge & language 52.9, Human preference 57.7
  1. 35Kimi K352.8
  2. 34Grok 4.753.0
  3. 35Qwen3.8-Max52.8

Published scores

BenchmarkScore
Terminal-Bench 2.1github.com88.3%
WeirdMLHåvard Tveit Ihle82.6%
LiveBench CodingLiveBench81.4%
FrontierSWEgithub.com81.2%
ProgramBenchgithub.com77.8%
Kimi Code Bench 2.0github.com72.9%
DeepSWE v1.1Datacurve68.5%
DeepSWE v1.1 (Kimi Code)github.com67.5%
SciCodegithub.com58.7%
MirrorCodeEpoch AI54.4%
MLS-Bench-Litegithub.com48.3%
SWE-Marathongithub.com42%
PostTrainBenchgithub.com36.6%
FrontierSWE V2Proximal Labs25.9%
Terminal-Bench 4.0Vals AI17.2%

Compare Kimi K3: Kimi K3 vs Kimi K2.6, Kimi K3 vs GPT-6.1 Sol, Kimi K3 vs Kimi K3.1 Pro, Kimi K3 vs Kimi K2.7 Code, Kimi K3 vs Kimi K2.5, Kimi K3 vs Kimi K2

Apps

Use Kimi K3 in the tools you already run. Each guide points the app at the Respan gateway.

Setup guides for more apps are in the integration docs.

Kimi K3 FAQ

Short answers, built from the data on this page.

Kimi K3 is a model from Moonshot, released on Jul 16, 2026. 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 navigating large repositories, using tools, debugging, and iterating against images, logs, tests, and runtime feedback. Its architecture uses KDA and Attention Residuals for computational efficiency.

On the Respan gateway, input starts at $0.192 per 1M tokens on OpenRouter, and output starts at $9.50 per 1M tokens on Chutes, after gateway discounts. Averaged across its 23 routes from 22 providers, it costs $2.46 per 1M input tokens and $14.95 per 1M output tokens.

Kimi K3 is available through 23 routes from 22 providers: Concentrate, Wafer, Makora, Chutes, Relace, MorphLLM, and 16 more.

Kimi K3 accepts up to 1M tokens of context on its largest provider route.

Point any OpenAI-compatible SDK at https://gateway.respan.ai/v1 with your Respan API key and set the model to "moonshot/kimi-k3". To set it up with a coding agent, copy the agent setup prompt at the top of this page.

Kimi K3 scores 52.8 out of 100 on the Respan Index v1.1, ranking #35 of 176 models. By category: Coding 38.8, Agents 61.8, Reasoning 54.8, Knowledge & language 52.9, and Human preference 57.7. The Respan Index combines independent and lab-reported benchmark results into one score, using benchmark data collected by Respan.

Kimi K3 has 82 published benchmark scores, including Terminal-Bench 2.1 88.3%, DeepSearchQA (F1) 95%, and GPQA Diamond 93.5%. Each score links to where it was published; the data is collected by Respan.

Certified where it counts.

Built to meet the security and privacy standards that enterprise and healthcare teams require.

ISO 27001

The internationally recognized standard for information security management.

SOC 2

Secure, compliant management of your data across all of our systems.

GDPR

Operated under GDPR, the world's strictest standard for data privacy.

HIPAA

HIPAA compliant, with a BAA available for healthcare teams.

Works with your entire stack

Use Respan with your favorite frameworks and tools.

AnthropicAI21 LabsxAIVercel AI SDK
LangChainOpenRouterAssemblyAICursor
LiteLLMGroqBasetenMem0
CohereCogneeLlamaIndexFireworks
DeepSeekLinkupHaystackTogether AI
PerplexityInference.netPostHogRubyLLM
SlackMastraAzure OpenAIMistral
AWS BedrockMoonshotHyperspellBAML
ModaInstructorGoogle GeminiParasail
ReductoZapierAgnoNebius
OpenAINovita AIReplicateSuperagent

Built for AI agents.
Break less.
Ship more.