Local AI Advisor

Can MacBook Pro M4 Pro 48 GB run Recursive Flint (40B-A2.5B)?

48 GB unified · 273 GB/s memory bandwidth · Recursive Flint (40B-A2.5B) has 39.8B parameters (2.5B active per token).

Short answer

Yes, excellent for chat and comfortable or better for every workload we test.

Best quantization for chat: MXFP4 (20 GB of weights) on LM Studio. Generation: 90–120 tok/s, leaving about 18 GB free for your other apps.

How it rates for each workload

Same model and machine, different jobs. Agents and long documents send far bigger prompts than chat, so they need much more speed and memory.

WorkloadVerdictQuantGenerationFirst replyContext
Chatwith Open WebUIExcellentMXFP490–120 tok/sunder 1 s8KDetails
Coding questionswith ContinueExcellentMXFP485–110 tok/s≈4 s16KDetails
Coding in a repositorywith AiderComfortableMXFP470–95 tok/s≈15 s32KDetails
Agentic codingwith OpenCodeComfortableMXFP465–90 tok/s≈20 s32KDetails
Long documents (64K)with Open WebUIExcellentMXFP445–60 tok/s≈55 s64KDetails

Speeds are estimates calibrated against measured benchmarks on similar hardware. “First reply” is the time to read the workload’s opening prompt. How we calculate

Different tool, context size or workload?

These pages use typical settings. The full check models your exact setup.