A new inference engine named Saber, developed in Rust, has been released, targeting AMD R9700 GPUs. Saber supports various tensor parallelisms and specifically Qwen 3.8 27b and Flash next models. Benchmarks indicate that Saber outperforms existing engines on these models, achieving high token generation speeds. AI
IMPACT Offers improved inference speeds for local LLM deployments on AMD hardware.
RANK_REASON Release of a new inference engine for specific hardware.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
