Models · Text

vikasit-30b-moe

MoE efficiency — 30B quality at 3B inference cost. Fast and smart.

Parameters

30B (3B active)

Context window

256K

Best for

Efficient inference

Benchmarks

MMLU-Pro80.9%

GPQA Diamond73.4%

AIME 202585.0%

LiveCodeBench66.0%

Thinking mode, no tools where applicable. See full comparisons on the benchmarks page.

# via Vikasit Inference (OpenAI-compatible)

base_url = "https://api.vikasit.ai"

model = "vikasit-30b-moe"