FULL STORY
Nanbeige4.3-33B: Open-Source Launch and First Benchmarks
BOSS Zhipin's Nanbeige team open-sourced the Nanbeige4.3-33B reasoning model, which follow-up benchmarks found lower hallucination rates but higher token consumption.
2026-11-06 ~ 2026-11-06 · 2 episodes · 4 posts
Episode 1 · BOSS Zhipin's Nanbeige Releases Open-Source 33B Model (2026-11-06, 2 posts)
Nanbeige, the foundation model team backed by BOSS Zhipin, released Nanbeige4.3-33B, a dense reasoning model with Apache 2.0 open weights that matches Qwen3.8 27B on Artificial Analysis' Intelligence Index.
- BOSS Zhipin's Nanbeige4.3-33B open model matches Qwen3.8 27B on AA Index — ArtificialAnlys · 2026-11-06
- BOSS Zhipin's Nanbeige releases open-weights 33B model, ties Qwen3.8 27B at 20 — ArtificialAnlys · 2026-11-06
Episode 2 · Nanbeige4.3-33B Shows Lower Hallucination but Higher Token Use (2026-11-06, 2 posts)
Artificial Analysis benchmarks show Nanbeige4.3-33B has a lower 18% hallucination rate than Qwen3.8 27B but fewer correct answers, and uses 158K tokens per task, 44% more than Qwen, despite tying at 20 on the Intelligence Index v5.
- Nanbeige4.3-33B hallucinates less than Qwen3.8 27B but answers fewer correctly — ArtificialAnlys · 2026-11-06
- Tied at 20, Nanbeige4.3-33B burns 158K output tokens per task, 44% more than Qwen — ArtificialAnlys · 2026-11-06