AI Models & Performance Benchmarks
One OpenAI-compatible endpoint with automatic prompt caching. Real model pricing, billed per token.
| MODEL | CONTEXT | INPUT / 1M | CACHED | OUTPUT / 1M | SPEED (TPS) | LATENCY |
|---|---|---|---|---|---|---|
GLM-5.3-FlashPOPULAR Zhipu AI (zai-org) · FP8 | 1M | $0.10 | $0.014 | $0.18 | ~280 t/s | ~0.25s |
DeepSeek-V4-FlashPOPULAR DeepSeek AI · FP8 MoE | 1M / 384K | $0.10 | $0.014 | $0.18 | ~105-250 t/s | ~0.45s |
DeepSeek-V4-Flash-0731 (Fast)POPULAR Fusion Low-Latency Tier · FP8 MoE | 1M / 384K | $0.12 | $0.016 | $0.22 | ~130-400 t/s | ~0.35s |
MiniMax-M2.7 MiniMaxAI · FP8 MoE | 204K / 131K | $0.25 | $0.02 | $1.00 | ~180 t/s | ~0.45s |
Ready to connect to Fusion Gateway?
Get your unified Fusion API key in seconds with local MMK payment support.