◆ Models

Top AI models, one gateway

Mix and match flagship and fast models. Switch with a single string change, no SDK swap needed.

Deepseek V4 Flash

Deepseek
fast

Lightweight Deepseek variant. Fast, cheap, and capable for high-volume tasks.

In $0.17Out $0.34Speed 198 t/s
128k context·8k output

Mimo V2.5

Xiaomi
fast

Xiaomi's fast and lightweight model. Great for chat, classification, and everyday tasks.

In $0.10Out $0.20Speed 180 t/s
128k context·8k output

Nemotron 3 Ultra 550B

NVIDIA
standard

NVIDIA's large open model. Strong reasoning and instruction following at scale.

In $0.25Out $0.50Speed 85 t/s
128k context·8k output

Gemma 4 31B

Google
standard

Google's Gemma 4 model. Balanced performance for multilingual and coding tasks.

In $0.20Out $0.40Speed 110 t/s
128k context·8k output