GLM 5.3 Flash (380B, local 8-bit) praised for one-shot note app beating frontier models
9r4n4y · reddit · 2026-09-17
A user reports building their long-wanted note app in one shot with GLM 5.3 Flash running locally (8-bit quant, vLLM, 380B parameters), claiming the output looked better than Gemini 3.8's attempt at the same app and even beat Opus 4.6 Adaptive Max, with very few bugs. They also used Qwen 3.6 35B A3B for online research. Subjective hands-on impressions only — no screenshots or benchmarks — so treat with skepticism.
More from Models
- GPT-6 Astra tops Terminal-Bench 4.0 at 57.7%, Claude Opus 5 hits 51.8% at half the price — shensi · 2026-09-17
- Dev take: models are smart enough now — focus on making them fail less — rickasaurus · 2026-09-17
- OpenAI Publishes Misalignment Disclosure Framework, Plus Six Incident Reports From Six Months of Training — Thom_Wolf · 2026-09-17
- A 'Quadrillion-Parameter' Model Surfaces on X, But Details Remain Unverified — KyeGomezB · 2026-09-17
- Sentence Transformers SparseEncoder: Sparse Retrieval in a Few Lines — tomaarsen · 2026-09-17
- Tencent releases open-source Hunyuan HY4 Preview with 770B params, 1M+ token context — anthara_ai · 2026-09-17