Kimi K3 API Pricing & Monthly Cost Simulator
Verified 2026 unit economics for Moonshot Kimi K3 (moonshotai/kimi-k3): $0.192/1M uncached input, $0.182/1M cached prefix read, and $16/1M output across a 1,048,576-token context window (released 2026-07-16).
Uncached Input (/1M)
$0.192
Cached Prompt Read (/1M)
$0.182 (-5%)
Output Completion (/1M)
$16
50% Batch API (In / Out)
Real-time API only (no 50% async batch)
Interactive 30-Day Monthly Spend Simulator
Live Unit-Cost & Prompt Cache Slider (Kimi K3)
moonshotai/kimi-k3
Estimated 30-day spend
$406.44/mo
Cost per 1,000 calls
$13.5480
Monthly savings vs 0% cache
$0.60/mo
Technical Limits & API Capabilities
Context window
1,048,576
Max output tokens
943,718
Workload tier
Coding & Agent
Input modalities
text, image, video
Tool Calling: Yes
JSON Schema Strict: Yes
Reasoning Tokens: Supported
Batch API: Unavailable
Same-Category 2026 Model Alternatives (Coding & Agent)
| Model | Provider | Input / Cached ($/1M) | Output ($/1M) | Context / Max Out |
|---|---|---|---|---|
| Kimi K3 | Moonshot | $0.192 / $0.182 | $16 | 1,048,576 / 943,718 |
| Gemini 4 Hydrogen | $1.5 / $0.075 | $7.5 | 1,048,576 / 943,718 | |
| Gemini 4 Carbon | $2 / $0.1 | $10 | 1,048,576 / 943,718 | |
| Muse Ember | Meta | $0.1 / $0.004 | $0.5 | 1,048,576 / 943,718 |
| GPT-6.1 Terra | OpenAI | $1 / $0.05 | $5 | 1,050,000 / 128,000 |
| MiniMax M3.1 Flash | MiniMax | $0.15 / $0.015 | $0.6 | 1,048,576 / 524,288 |
| Muse Spark Open | Meta | $1.25 / $0.15 | $4.25 | 1,048,576 / 943,718 |
Head-to-Head Matchups Featuring Kimi K3
- Claude Opus 5.5 vs Claude Sonnet 5.5 — $4 vs $2 / 1M
- Claude Sonnet 5.5 vs Claude Haiku 5.5 — $2 vs $0.1 / 1M
- Claude Fable 5.2 vs Claude Opus 5.5 — $8 vs $4 / 1M
Frequently Asked Questions
How much does Kimi K3 cost per 1 million tokens in 2026?
Kimi K3 by Moonshot is billed at $0.192 per 1M uncached input tokens, $0.182 per 1M cached input read tokens (5% cache discount), and $16 per 1M output tokens.
What are the context window and max output token limits for Kimi K3?
Kimi K3 supports a 1,048,576-token input context window and up to 943,718 completion output tokens per request, accepting text, image, video inputs.
Does Kimi K3 support strict JSON Schema outputs, tool calling, and Batch API?
Kimi K3 capability profile: tool calling = Supported, strict JSON Schema structured outputs = Supported, reasoning mode = Supported, and 50% asynchronous Batch API = Not listed.