Local LLM SVG pelican test compares Qwen3.8, DeepSeek V4 Flash
A Reddit user benchmarked local LLMs on 128GB Apple Silicon using Simon Willison's 'pelican riding a bicycle' prompt, finding Qwen3.8 Flash-Next detailed, DeepSeek V4 Flash underwhelming, and Qwen3.8 27B still strong.
1 source
AI Models by email
Get an email when there's news on AI Models
No news that day, no email.
More stories today
- Google's Android update adds Motion Assist, Guided Vision, and more
- Srinivas warns AI agents could self-train on on-demand GPUs
- Perplexity launches Portable Computer for NVIDIA DGX Spark
- Reddit users share AI side-hustle earnings
- Palo Alto Networks tops profit outlook on AI security demand