Local AI Advisor

Compare models on your machine

Not generic leaderboards: every model is rated for your hardware, your tool and your workload — so you can see which one will actually feel better to use.

Hardware
Use case
Tool
Other apps
Context
Phi-4 (14B)Recursive Flint (40B-A2.5B)Qwen3.6 35B-A3BQwen3.5 35B-A3B

MacBook Pro M4 Pro 48 GB · Agentic coding / medium repository

Phi-4 (14B)

5-bit · LM Studio

Acceptable

Usable for agentic coding, but expect compromises: generation (13–17 tok/s) is slow for agentic coding — slow steps compound across many agent iterations; ingesting a prompt of 14K tokens takes ≈85 s and each step ≈45 s; across ~25 calls a task takes ≈20 min.

MemoryExcellent

≈23 GB headroom

GenerationPoor

13–17 tok/s

Prompt processingPoor

≈85 s first prompt

ContextInadequate

16K context

Tool compatibilityExcellent

Official

Workload suitabilityPoor

Poorly suited

Recursive Flint (40B-A2.5B)

MXFP4 · LM Studio

Comfortable

Yes — this should be comfortable for agentic coding.

MemoryExcellent

≈15 GB headroom

GenerationExcellent

65–90 tok/s

Prompt processingGood

≈20 s first prompt

ContextExcellent

32K context

Tool compatibilityExcellent

Official

Workload suitabilityGood

Suited

Qwen3.6 35B-A3B

5-bit · LM Studio

Comfortable

Yes — this should be comfortable for agentic coding.

MemoryGood

≈12 GB headroom

GenerationExcellent

50–70 tok/s

Prompt processingGood

≈20 s first prompt

ContextExcellent

32K context

Tool compatibilityExcellent

Official

Workload suitabilityGood

Suited

Qwen3.5 35B-A3B

5-bit · LM Studio

Comfortable

Yes — this should be comfortable for agentic coding.

MemoryGood

≈12 GB headroom

GenerationExcellent

50–70 tok/s

Prompt processingGood

≈20 s first prompt

ContextExcellent

32K context

Tool compatibilityExcellent

Official

Workload suitabilityGood

Suited