A user benchmarked the IFM/K2-Horizon-7B model on a system with 16GB of VRAM, finding it significantly underperformed compared to Qwen3.8-27B and Ornith-1.5-9B. Despite fitting the model entirely within the 16GB VRAM, the K2-Horizon-7B model struggled, with several tasks timing out and one resulting in broken code. The benchmark indicated that Qwen3.8-27B was the top performer, even with a less optimal quantization, while Ornith-1.5-9B offered a good balance of speed and usability. AI
IMPACT Highlights performance limitations of smaller models on consumer hardware and reinforces the strength of larger, established models.
RANK_REASON User benchmark of a specific model's performance on limited hardware. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
