Open Models, Grounded in Law

You don't need to build custom legal models from scratch. See how current open-weight foundation models trade guesswork for verifiable citations when grounded by Dike's official MENA legal layer.

Filter:

Flagship Engine30 Billion•262k Tokens (extensible to 1M) Context•Apache 2.0

Qwen3.6 30B

Alibaba's flagship-coding dense model, grounded in Dike's official MENA legal index.

Citation Accuracy:

Raw: Invents Articles 87-90 with fabricated sub-rules+ Dike Data: Quotes the real Article 107 verbatim

View Grounding Benchmarks

31 BillionApache 2.0

Gemma 4 31B

Google's multimodal dense model, grounded for bilingual contract review and cross-jurisdiction comparison.

Citation Accuracy:Quotes the actual article pair from each corpus

Explore model

30 BillionApache 2.0

Muse Glimmer 30B

Meta's on-device agentic model, grounded for local, low-latency GCC compliance workflows.

Citation Accuracy:Cites Article 85, correct 2/3 fraction

Explore model

30B total (3B active, MoE)OpenMDW 1.1

Nemotron 3.5 Lightning

NVIDIA's fast hybrid MoE agent model, grounded in Dike's official MENA legal index.

Citation Accuracy:Correct: 90 days from award notification

Explore model

30 BillionApache 2.0

Granite 4.1 30B

IBM's enterprise hybrid Mamba-Transformer model, grounded in Dike's official MENA legal index.

Citation Accuracy:Cites Articles 19 and 20 verbatim

Explore model