Cheaper Inference
Model record / Moonshot
kimi-k3
Access kimi-k3 from Moonshot through one OpenAI-compatible API with live market-linked pricing and usage-based billing.
Text generationVision inputReasoningStreaming35% below list
What you pay
Current catalog rates.
Input / 1M tokens$3.00 list$1.95
Output / 1M tokens$15.00 list$9.75
Cached input / 1M tokens$0.30 list$0.195
Current discount35%
Observed reliability
99.54
Successful provider-reaching requests
Output speed
45.7
Median tokens per second over sustained generations
TTFT
3.8 s
p50 · p90 12.4 s · n 8,210
Latency
7.9 s
p50 · p90 32.7 s · n 9,967
Evidence strength
—
At a glance
- Provider
- Moonshot
- Model ID
- kimi-k3
- Type
- Text generation
- Endpoints
- /v1/chat/completions, /v1/responses, /v1/messages
- Input
- image, pdf, text, video
- Output
- text
- Vision input
- Supported
- Image editing
- Not supported
- Video input
- Not supported
- Reasoning
- Supported
- Streaming
- Supported
- Context window
- 1,000,000 tokens
Parameters by endpoint
The request field of each parameter on each endpoint, and the values it takes.
POST /v1/chat/completions
| Field | Values |
|---|---|
| reasoning.exclude | |
| max_tokens | |
| reasoning | effort: low, high, max |
| response_format | |
| stop | |
| tool_choice | |
| tools |
POST /v1/messages
| Field | Values |
|---|---|
| max_tokens | |
| output_config.format | |
| stop_sequences |
Pricing history
See how much this model saves and when its recorded price has been lowest.
Time zone
kimi-k3
Growing confidenceThis model averaged 35.0% below comparable direct provider prices across the selected range. Its strongest observed window was 6:00 PM–4:00 PM in UTC.
Last 24 hours
Average saved
35.0%
Last 24 hours
Cheapest time
6:00 PM–4:00 PM
35.0% average savings
Hourly coverage
23/24
23 total observations
Signal quality
Growing
Ready for planning
Typical unit savings$1.05 / 1M input tokens$5.25 / 1M output tokens
When this model is cheapest
Each bar is the average discount for that hour in UTC (UTC). Missing bars mean no observation was recorded.
ObservedCheapestNo data
Hover a bar for savings per token and observation counts.
Source buckets are UTC; labels are converted to UTC.
Last 24 hours
#35
of 408 ranked models
Last 7 days
#33
of 418 ranked models
Last 30 days
#21
of 418 ranked models
Ready when you are
Stop paying list price for tokens.
Marketplace pricing, capped at list, optimized further.