Kimi K3 (Global) pricing, context window and limits
Kimi K3 (Global) costs $3 per million input tokens and $15 per million output tokens at Amazon Bedrock list price. It accepts up to 1,048,576 tokens of context and returns up to 128,000 tokens.
- Model id
bedrock/global.moonshotai.kimi-k3 - Snapshot
What does Kimi K3 (Global) cost?
| Input | $3 |
|---|---|
| Output | $15 |
| Cached input (read) | $0.3 |
| Cache write | $3.75 |
US dollars per million tokens. A dash means no price is published.
How much can Kimi K3 (Global) read and write?
| Context window | 1,048,576 |
|---|---|
| Max output | 128,000 |
What can Kimi K3 (Global) do?
| Vision (image input) | Yes |
|---|---|
| Tool calling | Yes |
| Structured output | Yes |
| Reasoning | Yes |
| Input | text, image |
| Output | text |
What does a request to Kimi K3 (Global) cost?
| A typical request: 1,000 tokens in, 500 out | $0.011 |
|---|---|
| A long request: 10,000 tokens in, 1,000 out | $0.045 |
| 1,000 typical requests | $10.5 |
| 1 million tokens in and 1 million out | $18 |
Input read from the provider's prompt cache costs $0.3 per million tokens instead of $3, 90% less. A request that repeats a 1,000-token prompt from the cache costs $0.0078 instead of $0.011.
Among 199 Amazon Bedrock models with a published price, Kimi K3 (Global) is the 132nd cheapest by input price. Its context window is the 26th largest of 199. List prices only; discounts, batch prices and prices above 200K tokens are not shown.
How do I call Kimi K3 (Global) through DVARA?
Configure the bedrock provider in the DVARA LLM Gateway and send the model id bedrock/global.moonshotai.kimi-k3. DVARA, the AI gateway, applies your policies, audit and cost attribution to the call. See LLM cost attribution and budgets.
In DVARA, if you haven't set your own price for a model, its calls are costed at the price in the catalogue that ships with your DVARA version. A price you set in Flightdeck always wins. You can switch catalogue prices off. Then a model with no price you set gets no cost record: its calls count against no budget and appear in no report, and the gateway records the model as unpriced. Every cost record says which price it used.
Which other Amazon Bedrock models are there?
- Kimi K2 Thinking$0.6 in · $2.5 out · 262K context
- Kimi K2.5$0.6 in · $3 out · 262K context
- Kimi K3 (US)$3.3 in · $16.5 out · 1.05M context
- Claude Sonnet 5.5 (Global)$2 in · $10 out · 1M context
- GPT-5.6 Luna (Global)$0.2 in · $1.2 out · 1.05M context
- Claude Sonnet 5 (Global)$2 in · $10 out · 1M context
- GPT-5.6 Sol (Global)$4 in · $20 out · 1.05M context
- Claude Sonnet 4.6 (Global)$3 in · $15 out · 1M context
- GPT-5.6 Terra (Global)$2 in · $12 out · 1.05M context
- Claude Sonnet 4.5 (Global)$3 in · $15 out · 200K context
- GPT-6 Astra (Global)$10 in · $50 out · 1.05M context
- Claude Sonnet 4 (Global)$3 in · $15 out · 200K context
Which models from other providers are similar to Kimi K3 (Global)?
These are the closest in context window and input price, one per provider.
- Kimi K3Moonshot AI · $3 in · $15 out · 1.05M context
- Claude Sonnet 4.6Anthropic · $3 in · $15 out · 1M context
- GPT-5.4OpenAI · $2.5 in · $15 out · 1.05M context
- Qwen3.7 MaxAlibaba Qwen · $2.5 in · $7.5 out · 1M context
- Gemini 3.1 Pro PreviewGoogle Gemini · $2 in · $12 out · 1.05M context