2026 local model
getting-started & selection guide
Start from your device and real tasks, not from a parameter leaderboard.
Start reading ↗$ local-ai check --device✓ 32GB unified memory✓ local inference ready
OPEN COMMUNITY · BUILDING IN PUBLIC
Run AI
on your own device
An open technical community for local, on-device, edge and private AI. Learn the tech, choose models, reproduce benchmarks, join projects and solve real problems together.
48Articles
31Project profiles
12Open benchmarks
05Open bounties
LOCAL STACK / 01LIVE
MODELSYOUR
DEVICEPRIVATE
DATAREAL
APPS
DEVICE OWNEDOFFLINE READYOPEN SOURCE
DEVICE ADVISOR · ALPHA
What models can my device run?
Pick your setup and use case to get a practical local AI starting point.
RECOMMENDED STACKMatch 86%
01
Recommended model14B 级开放模型 · Q4_K_M
EngineMLX / Ollama
Est. speed≈ 22–45 tok/s
Great for daily chat, writing and Q&A. Actual performance depends on model version, context length and system environment.
COMMUNITY RADAR
Worth watching this week
2026 local model
Apple M4 Pro:
12 models tested
Complete records of speed, memory, long context and Chinese tasks.
42.8tok/s
BEST RESULT
BEST RESULT
Domestic AI PC
open benchmark bounty
Build the first open, reproducible inference benchmark.
Reward¥8,000
START YOUR PATHWherever you start,
Wherever you start,
there's a next step.
Learn, reproduce, contribute, launch — the community offers clear entry points for every stage.