viraldi.
Back to discovery
10:10
Technology & Computing / VIDEO

Nanbeige4.3-33B Is CRAZY Good (On Paper): I Tested It on a Rented GPU

Prompt Engineer 48· 30.3K subscribers

12d ago · English audio

Nanbeige4.3-33B is an Apache 2.0, 33B open model that the lab says beats Qwen3.8-27B on Terminal-Bench 2.1 (80.9 vs 73.0). I rented a GPU, loaded it in vLLM, and tried to break it. Real logs, real failures, no invented numbers. Model card: https://huggingface.co/Nanbeige/Nanbeige4.3-33B RunPod (GPUs I used): https://get.runpod.io/pe48…

5.1×Score at discovery
3.4KViews
662Median · 20 videos

Outlier score

This score was measured on Sat, 07 Nov 2026 00:38:28 GMT. Current view counts may differ.

3.4K views ÷ a 662 baseline = 5.1×. Compared with 20 earlier videos from the same channel, matching the format recorded at scoring time. Their counts were observed at comparable video ages.

YouTube snapshot: Sat, 07 Nov 2026 21:05:39 GMT. Counters may have changed.

Read the methodology

Generative AI & Agents

Channel-balanced percentile: 36

Comparison details

166 videos from 113 other channels, with each channel weighted equally. Same format and language, with 10,000–99,999 subscribers. Observed at 305 hours after publication, within ±48 hours.

This describes the collected sample, not all of YouTube. Updated Sat, 07 Nov 2026 22:16:21 GMT.

Niches & tags

Transcripts and detailed analytics are available for recent outlier Shorts up to three minutes. Research tools