Eight open-weight models past 30B ship permissive licenses and GPQA claims in 2026 — spanning 31B to 2.8T parameters

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 live-web scan on 2026-09-01 found eight qualifying releases between 2026-02-24 and 2026-07-16. All carry Apache 2.0 or MIT licenses; the >30B cutoff uses total parameter count, not MoE active parameters, and GPQA figures are kept exactly as reported.

Apache 2.0MITCircle area — total parameters (log scale on the axis) · label shows reported GPQA figure

The reported GPQA range is narrow — 84.2 to 91.2 — while total parameter counts spread 90-fold, and the top score belongs to GLM 5.2 at 744B, not to the largest model. techjacksolutions.com

Five of the eight are mixture-of-experts designs whose active counts fall as low as 3B; both Qwen 35B releases qualify only on total parameters. awesomeagents.ai

Kimi K3, at 2.8T the largest model on the list, reports only an approximate "~87%" GPQA figure from a single supporting page. gist.github.com

Evidence depth varies: GLM 5.2 and Inkling each have 4 supporting pages found, while Gemma 4 31B, Qwen3.5 and Kimi K3 rest on one page apiece. enterprisedna.co

Model	Parameters (as reported)	License	Release	Reported GPQA	Pages	Source

Live-web scan conducted 2026-09-01 across broad web, model-card, repository, benchmark and model-specific searches. 8 rows, one per exact model release. "Permissive" means Apache 2.0 or MIT; "after January 2026" means after 2026-01-31; the >30B cutoff uses total parameters, not MoE active counts. GPQA figures are reported — often self-reported or repeated by secondary sources — and mix GPQA and GPQA-Diamond variants under differing protocols, so they are not directly comparable; approximate signs and labels are preserved verbatim. Comprehensive to discoverable indexed evidence, not a formal registry.

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.