# Best AI model under 32B for factuality — October 2026

As of Oct 15, 2026, Gemma 4 31B & Gemma 4 26B A4b & Qwen3 8 27B are tied for state of the art in Factuality, models at or under 32B (since Oct 8, 2026; sources LMArena Text factuality; methodology v1.0).

Category board: https://isitsota.com/c/factuality?slice=le-32b

## Top 10

- #1 Gemma 4 31B
- #1 Gemma 4 26B A4b
- #1 Qwen3 8 27B
- #2 Qwen3 5 27B
- #3 Granite 4.2 30B
- #4 Nvidia Nemotron 3.5 Lightning 30B A3b Nvfp4
- #5 Nvidia Nemotron 3 Nano 30B A3b Bf16
- #6 Granite 4.2 8B
- #7 Olmo 3.1 32B Instruct
- #8 Granite 4.2 3B

## Crown changes

- Oct 8, 2026 — Factuality: Gemma 4 31B and Gemma 4 26B A4b and Qwen3 8 27B are co-SOTA

## Weights

- SimpleQA Verified: primary, weight 1
- LMArena Text factuality: primary, weight 0.8