Best local LLMs by hardware tier
Rankings only make sense once you fix the hardware. Pick a tier below — the leaderboard re-sorts to the models the community actually runs there, weighted by report count, fastest observed tokens-per-second, and recency.
Looking for a coding model? See the best local LLM for coding by VRAM.
Apple Silicon 24–36 GB
Lower-tier Apple Silicon. 13B–30B with MLX or llama.cpp Metal. Ranked from 8 reports.
| # | Model family | Best variant tested | Reports | Fastest t/s |
|---|---|---|---|---|
| 1 | Qwen3.8Alibaba Flash-Next · 125B-A6B · Q4 · on M4 32GB | Flash-Next · 125B-A6B · Q4 on M4 32GB | 6 | 22.0 |
| 2 | DeepSeek V4 FlashDeepSeek 284B-A13B · 284B-A13B · 4bit · on M5 32GB | 284B-A13B · 284B-A13B · 4bit on M5 32GB | 1 | 1.0 |
| 3 | Gemma 4Google DeepMind — · on M5 32GB | — on M5 32GB | 1 | — |
How we rank
A single global "best models" list doesn't really exist — what runs well on a 5090 is often unrunnable on a 4060, and a 7B that screams on an M3 Max is usually a poor pick on an H100. So we fix the hardware first, then rank the families that actually have community reports on it. The score blends popularity (log-scaled report count), fastest observed tokens-per-second normalized within the bucket, recency (90-day half-life), and a small bias for rows where we know the variant + quant + GPU cleanly. Click into a family for the full breakdown of records.