Hailuo 2.3 (海螺视频)
MINIMAX:HAILUO-2.3Video FoundationBusyLoad:75%SLA:99.94%等待 0.8s
MiniMax Spatial-Temporal Autoregressive Diffusion
Base Rate
55 Credits / run
24h +16.75%
MiniMax cinematic text/image-to-video flagship renowned for subtle facial expressions and complex physical world dynamics.
24h Total Calls
1.42Mcalls
+16.75% vs yesterday
Total Generations
0.98Mvids
+15.2% · Expression 98%
Avg Latency (P50)
9.1s
5s Physics 9.1s · Expression 98%
Availability SLA
99.94%
Failover <105ms
Model Telemetry & Volume Chart
MINIMAX:HAILUO-2.3Invocations
62,160次+26.39%
Live Telemetry
Operational Performance
Production-grade Latency & Concurrency
P50 TTFT380 ms
P99 Latency15400 ms
Throughput TPS
Peak Concurrency720 req
Packet Loss Rate0.002%
Cost & Billing Matrix
Credits and Token Breakdown
Base Task Cost55 Credits
Input Cost / 1k Tokens¥0.018
Output Cost / 1k Tokens¥0.055
4K Upscale Multiplier1.8x
VIP Discount至高 25% 优惠
Technical Specifications
Parameters, Context & Resolutions
ArchitectureMiniMax Spatial-Temporal Autoregressive Diffusion
Parameters15 Billion
Context Window2,048 tokens
Max Resolution1080P Cinema Grade
Max Duration6s / 10s
Global Edge Gateways
Live Region Latency & Load Balancing
East China (Shanghai)
18ms
South China (Guangzhou)
22ms
Hong Kong Gateway
34ms
Asia Pacific (Singapore)
54ms
US West (Silicon Valley)
146ms
EU Central (Frankfurt)
170ms
Comparable Models in Same Domain
Models with similar creative modalities
Kling 2.6 Pro (可灵)
KUAISHOU:KLING-2.6-PRO
Kuaishou state-of-the-art video model with cinematic camera paths, high-motion human dynamics, and physical consistency.
Rate60 Credits / run
Google Veo 3
GOOGLE:VEO-3-GEN
DeepMind native audio-visual video model mastering cinematic framing, focus pulling, and fluid temporal coherence.
Rate75 Credits / run