AI21 Labs

weights

open(downloadable on HF, contact-info gate)

data

closed

code

partial(inference/usage only, no training pipeline)

license

Jamba-Open-Model-License(non-OSI: $50M annual-revenue commercial cap + AUP + naming/attribution requirements)

Weights are downloadable under the Jamba Open Model License, non-OSI and capped at $50M in annual revenue for commercial use, though the vendor's 'Open Model' branding overstates that restriction. The license bounds rather than forbids commercial use, unlike CC-BY-NC or Mistral's Non-Production License, which forbid it outright.

• https://huggingface.co/ai21labs/AI21-Jamba-1.5-Large recorded 2026-09-29

  Model card states "a permissive license allowing full research use and commercial use under the license terms"; repo tags include "license:other" and the repo is gated "auto"; the page is a usage card with inference examples, and searching the body for a training corpus, training data or training pipeline found none.

• https://www.ai21.com/licenses/jamba-open-model-license recorded 2026-09-29

  License text: "generate more than fifty million US Dollars (USD $50,000,000) in annual revenue" requires a separate license; a Notice file with attribution is also required.

Adoption is measured as Hugging Face downloads for AI21-Jamba-1.5-Large, the only repository declared on this product. Download volume is minimal, reflecting little ongoing use.

• https://huggingface.co/api/models/ai21labs/AI21-Jamba-1.5-Large recorded 2026-09-29

  166 downloads in the trailing 30 days for ai21labs/AI21-Jamba-1.5-Large; the API record reads "id":"ai21labs/AI21-Jamba-Large-1.5" and "downloads":166

Jamba 1.5 Large is a 2024 hybrid SSM-Transformer that led open models on long context at release but is now mid-tier against the 2026 frontier. The HF card gives it Arena Hard 65.4, MMLU (CoT) 81.2 and MMLU Pro (CoT) 53.5, ahead of Llama 3.1 70B and 405B at launch but well short of the current reasoning and coding frontier.

• https://huggingface.co/ai21labs/AI21-Jamba-1.5-Large recorded 2026-09-29

  Benchmark table, Jamba 1.5 Large: Arena Hard 65.4, MMLU (CoT) 81.2, MMLU Pro (CoT) 53.5; includes a "RULER Benchmark - Effective context length" section.

• https://www.ai21.com/blog/announcing-jamba-model-family/ recorded 2026-09-29

  Launch post: "Jamba 1.5 Large, with a score of 65.4, outpaces both Llama 3.1 70B and 405B" on Arena Hard.