MoE model G9v3-39A5B benchmarks top Qwen 3.6, seeking real-world tests
LegacyRemaster · reddit · 2026-08-20
A Reddit user highlights the G9v3-39A5B model, noting its strong performance on ArtificialAnalysis where it surpasses the previous leader Qwen 3.6 27B. The user points out that it is a Mixture-of-Experts (MoE) model and mentions a llama.cpp branch supporting it. The community is invited to share test results.
More from Models
- Claude Advances Math Bound While Attempting Riemann Hypothesis — PtrPomorski · 2026-08-20
- LLMs Struggle with Simple Anagram Task: Claude Succeeds, Others Fail — HydronautInSpace · 2026-08-20
- Rumor: Fable 5.1, GPT-6, and Other Major Models Incoming — bindureddy · 2026-08-20
- Developer praises Grok 4.6, abandons other models — elonmusk · 2026-08-20
- Claude reportedly refuses to work with 90% of weekly quota still remaining — MoreFaithlessness954 · 2026-08-20
- 7 parameters that control every LLM response explained — blaizedsouza · 2026-08-20