Loading
Loading
Model record / MiniMax
minimax-m3
Access minimax-m3 from MiniMax through one OpenAI-compatible API with live market-linked pricing and usage-based billing.
Text generationVision inputStandard generationStreaming35% below list
What you pay
Current catalog rates.
Input / 1M tokens$0.30 list$0.195
Output / 1M tokens$1.20 list$0.78
Cached input / 1M tokens$0.06 list$0.039
Current discount35%
Requests above 512,000 input tokens use $0.39 input and $1.56 output per 1M tokens.
Observed reliability
99.77
Successful provider-reaching requests
Output speed
182.4
Median tokens per second over sustained generations
p95 first token (ms)
6,519
p50 3491 ms · total p95 12659 ms
Evidence strength
—
established
At a glance
- Provider
- MiniMax
- Model ID
- minimax-m3
- Type
- Text generation
- Endpoint
- /v1/chat/completions
- Vision input
- Supported
- Image editing
- Not supported
- Video input
- Not supported
- Reasoning
- Not supported
- Streaming
- Supported
- Context window
- 200,000 tokens
- Maximum output
- 512,000 tokens
Send a request
Pricing history
See how much this model saves and when its recorded price has been lowest.
Time zone
minimax-m3
Growing confidenceThis model averaged 35.0% below comparable direct provider prices across the selected range. Its strongest observed window was 1:00 AM–11:00 PM in UTC.
Last 24 hours
Average saved
35.0%
Last 24 hours
Cheapest time
1:00 AM–11:00 PM
35.0% average savings
Hourly coverage
23/24
23 total observations
Signal quality
Growing
Ready for planning
Typical unit savings$0.11 / 1M input tokens$0.42 / 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
#19
of 93 ranked models
Last 7 days
#24
of 94 ranked models
Last 30 days
#38
of 96 ranked models
Ready when you are
Stop paying list price for tokens.
Marketplace pricing, capped at list, optimized further.
