llm-bench
.io
— Local LLM Leaderboard
Real-world benchmarks from actual users — LLM-judged
quality
and
speed
across coding, agent, role-play & research tasks, grouped by VRAM.
Run your own test →
__META__
Model
VRAM
Runs
Quality
Speed (tok/s)
Coding
Agent
Role-Play
Research
Updated
__ROWS__