Models · Text

vikasit-120b

Datacenter MoE. Frontier reasoning at low inference cost — only ~5B active per token.

Parameters

120B (5B active)

Context window

262K

Best for

Datacenter

Benchmarks

MMLU90.0%

MMLU-Pro80.8%

GPQA Diamond80.1%

AIME 202495.8%

AIME 202592.5%

SWE-bench Verified62.4%

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-120b"