Thirteen open-weight models over 30B carry a true MIT or Apache 2.0 license — and GLM-5.2 tops their reported GPQA at 91.2

Asked:

“Which open-weight AI models have a permissive license, more than 30B parameters, a release date after January 2026, and a reported GPQA score? List all four attributes for each model.”

A web scan on 2026-09-04 surfaced 23 candidate model records; after alias deduplication and a strict license test (standard MIT or Apache 2.0 only), 13 unique releases from February to August 2026 qualify, spanning 31B to 1.6T total parameters with reported GPQA Diamond scores of 71.2–91.2.

MIT Apache 2.0 GPQA score unverified — total parameters (log scale) vs reported GPQA, bubble hover for detail

The smallest qualifier, Gemma 4 31B Dense (Apache 2.0), reports 84.3 GPQA Diamond with the widest source support — 23 pages including infoq.com.

DeepSeek V4 Pro is the largest strict-license model at 1.6T total parameters under plain MIT, reported at 90.1 GPQA Diamond — chatforest.com.

Kimi K2.6 (90.5) and Kimi K3 (93.5) would beat most of the list but ship under Modified MIT / custom terms, so they fail the strict test — contextstudios.ai.

Four rows rest on a single source — Nex-N2-Pro, Inkling, GLM-5.3-Flash and GLM-5 — and GLM-5.3-Flash's extracted score (912.0) is a decimal error left unverified — dataconomy.com.

All 13 strict-license matches

Model	Release	Params (B, total)	License	GPQA (reported)	Variant	Sources	Link

Method: web scan performed 2026-09-04, merging broad discovery with targeted model-card and official-source searches; 23 raw candidate rows reduced to 13 unique releases after alias deduplication (Qwen 3.5 = Qwen3.5-397B-A17B; all Gemma 4 31B aliases merged; generic 1.6T DeepSeek V4 merged into V4 Pro; DeepSeek V4 "Flash Max" merged into Flash). "Permissive" means standard MIT or Apache 2.0 only: Kimi K2.6/K3 (Modified MIT/custom), NVIDIA Nemotron (OpenMDW / Nemotron Open Model License), MiniMax M3 (Community license) and rows labeled only "Open Weights" were excluded. Parameters are total, not active, for MoE models. GPQA scores are as reported by sources, not independently reproduced; variants shown because scores are not comparable across variants. GLM-5 release date disputed (Feb 11 vs 17; earliest shown); GLM-5.1 varies Apr 7–8 (earliest shown). GLM-5.3-Flash's extracted GPQA (912.0) is a decimal extraction error and is flagged as needing verification.

This report was generated automatically by Keenable SELECT at a user's request, from publicly available web sources linked herein. Keenable does not review, verify, or endorse its contents and makes no representation as to accuracy, completeness, or timeliness; AI-based extraction may contain errors. Nothing in this report is investment, legal, financial, or other professional advice. All trademarks and referenced content remain the property of their respective owners; no affiliation or endorsement is implied. To report an error, rights concern, or request removal: legal@keenable.ai.