Model family

INTELLECT

Family overview — summarizes releases; licenses and availability belong to each release.

INTELLECT is Prime Intellect's series of open-weight models. INTELLECT-1 is a 10B model trained from scratch on 1 trillion tokens with compute from distributed community contributors; INTELLECT-2 is a 32B model produced by a distributed reinforcement-learning run on QwQ-32B; INTELLECT-3 (November 2025) is a 106B mixture-of-experts model (12B active) post-trained from GLM-4.5-Air-Base with supervised fine-tuning and large-scale RL; and INTELLECT-3.1 continues INTELLECT-3 with further RL.1 2 3 4 5

Release notes: INTELLECT-3 announcement (external site: primeintellect.ai)
Paper: INTELLECT-3 technical report (arXiv 2512.16144) (external site: arxiv.org)
Repository: prime-rl repository (external site: github.com)

Last reviewedSep 29, 2026Entry updated Sep 29, 2026Documented release Unknown

Releases assessed

Each release is assessed on its own. This family has no family-wide openness label.

Releases in the INTELLECT family
Release	Weights	Tier (USASI rubric v0.1)	License	Released
INTELLECT-3	Weights: Public	Model-disclosure tier (USASI rubric v0.1): Open-weight	MIT, Apache-2.0, MIT	Nov 26, 2025
INTELLECT-3.1	Weights: Public	Model-disclosure tier (USASI rubric v0.1): Open-weight	MIT, Apache-2.0, MIT	Unknown

What it is useful for

The recent INTELLECT models are reasoning models aimed at math, code, science, and agentic tasks such as tool use and software engineering, and Prime Intellect uses them to demonstrate its open training stack (prime-rl, verifiers, and the Environments Hub).5 6 4

Organization context

Prime Intellect

Provenance and derivatives

INTELLECT-1 was trained from scratch by Prime Intellect. INTELLECT-2 was trained from Qwen's QwQ-32B, and INTELLECT-3 and INTELLECT-3.1 from GLM-4.5-Air-Base, a base model released by Z.ai (zai-org). Those base models are not catalog members and are not treated as U.S.-developed.1 2 3 7

Derived from: QwQ-32B (external site: huggingface.co) — Base model for INTELLECT-2; not a catalog member.
Derived from: GLM-4.5-Air-Base (external site: huggingface.co) — Base model for INTELLECT-3 and INTELLECT-3.1; not a catalog member.

U.S. eligibility

Project eligibility rests on documented governing or maintaining entities, not on contributors.

Eligible · basis: U.S. headquarters

The INTELLECT models are trained and published by Prime Intellect, Inc., a Delaware corporation with U.S. locations (see the prime-intellect record). For releases built on third-party base models, eligibility rests on Prime Intellect's own training and does not extend to the base models. Each release is assessed on its own record.3 8

Assessed Sep 29, 2026

Sources

1.
PrimeIntellect/INTELLECT-1 model card (external site: huggingface.co)
Prime Intellect · Model card · accessed Sep 29, 2026
2.
PrimeIntellect/INTELLECT-2 model card (external site: huggingface.co)
Prime Intellect · Model card · accessed Sep 29, 2026
3.
PrimeIntellect/INTELLECT-3 model card (external site: huggingface.co)
Prime Intellect · Model card · accessed Sep 29, 2026
4.
PrimeIntellect/INTELLECT-3.1 model card (external site: huggingface.co)
Prime Intellect · Model card · accessed Sep 29, 2026
5.
INTELLECT-3: A 100B+ MoE trained with large-scale RL (external site: primeintellect.ai)
Prime Intellect · Announcement · published Nov 26, 2025 · accessed Sep 29, 2026
6.
INTELLECT-3: Technical Report (arXiv 2512.16144) (external site: arxiv.org)
Prime Intellect · Paper · published Dec 18, 2025 · accessed Sep 29, 2026
7.
zai-org/GLM-4.5-Air-Base model card (external site: huggingface.co)
Z.ai · Model card · accessed Sep 29, 2026
8.
Prime Intellect, Inc. Form D (accession 0001231919-26-000720) (external site: sec.gov)
U.S. Securities and Exchange Commission · Filing · published Jul 1, 2026 · accessed Sep 29, 2026

Report a correction Edit this entry

This listing is not an endorsement, a safety assessment, or a federal approval.