INTELLECT-3.1
Release in the INTELLECT family · version INTELLECT-3.1
Maintained by Prime Intellect1
A 106B-parameter mixture-of-experts reasoning model (12B active) that continues the training of INTELLECT-3 with additional reinforcement learning on math, coding, software-engineering, and agentic tasks, using prime-rl and environments built with the verifiers library.1
- Model hub: Model card (Hugging Face) (external site: huggingface.co)
- Paper: INTELLECT-3 technical report (arXiv 2512.16144) (external site: arxiv.org)
- Repository: prime-rl (training framework) (external site: github.com)
Availability and license
Overall availability
Weights download from Hugging Face without an access gate.12
Availability is separate from permission: read the license before using or redistributing.
Apache License 2.0 (prime-rl) (external site: raw.githubusercontent.com)5
MIT License (verifiers) (external site: raw.githubusercontent.com)6
The MIT license for the weights is declared in the model card metadata; the repository has no separate LICENSE file. The card says the model, training frameworks, and environments are released under MIT and Apache 2.0 licenses.13
Model-disclosure tier
The model parameters for this release can be downloaded by the public. License terms may still restrict use, redistribution, or commercial use.
Public materials checklist
| Item | Status | Notes and evidence |
|---|---|---|
| WeightsCan the general public download the model parameters for this release? | Public | BF16 safetensors in an ungated Hugging Face repository.12 |
| Inference codeIs code for running the model published? | Public | The model card gives a vLLM serving command with the qwen3_coder tool-call parser and deepseek_r1 reasoning parser.1 |
| Training codeIs the code used to train the model published? | Partial | The card says training was done with prime-rl (Apache 2.0) and verifiers environments (MIT), both open source. This covers Prime Intellect's training only, not the pretraining of the GLM-4.5-Air base.156 |
| Training-data informationPublic = the training data itself can be obtained. Partial = composition or sources are documented without full access. | Partial | The card names the task areas of the added RL (math, coding, software engineering, agentic tasks) and says all training and evaluation environments are on the Environments Hub, but does not list which environments or datasets were used for 3.1. The INTELLECT-3 data it builds on is described in the INTELLECT-3 report.14 |
| Training recipeAre the training configuration and procedure documented in enough detail to follow? | Partial | The card describes 3.1 only as continued RL from INTELLECT-3 and refers to the INTELLECT-3 technical report, which documents the INTELLECT-3 post-training; settings for the additional 3.1 RL are not given.14 |
| Evaluation materialsPublic = evaluation code or prompts that let others re-run the evaluations are published. Partial = results only. | Unknown | The INTELLECT-3.1 model card does not report evaluation results, and no separate 3.1 evaluation report was found. |
What it is useful for
The model card presents it as an update of INTELLECT-3 aimed at math, coding, software-engineering, and agentic tasks.1
Run and use notes
Organization context
Provenance and derivatives
Continued training of Prime Intellect's INTELLECT-3, which was itself post-trained from GLM-4.5-Air-Base, a base model released by Z.ai (zai-org). The base model is not a catalog member and is not treated as U.S.-developed.17
- Derived from: INTELLECT-3 — Same organization; starting checkpoint.
- Derived from: GLM-4.5-Air-Base (Z.ai) (external site: huggingface.co) — Third-party base model underlying INTELLECT-3; not a catalog member.
Other releases in the INTELLECT family
- INTELLECT-3Model-disclosure tier (USASI rubric v0.1): Open-weight
U.S. eligibility
Eligible · basis: U.S. headquarters
Trained and published by Prime Intellect, Inc., a Delaware corporation with U.S. locations (see the prime-intellect record). Eligibility covers Prime Intellect's training only; the underlying GLM-4.5-Air-Base model was released by Z.ai and is not treated as a U.S.-developed model.187
Sources
This listing is not an endorsement, a safety assessment, or a federal approval.