Local AI Advisor

Can Mac Studio M4 Max 64 GB run Recursive Flint (40B-A2.5B)?

64 GB unified · 546 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: 170–230 tok/s, leaving about 33 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 WebUIExcellentMXFP4170–230 tok/sunder 1 s8KDetails
Coding questionswith ContinueExcellentMXFP4160–210 tok/s≈2 s16KDetails
Coding in a repositorywith AiderComfortableMXFP4130–180 tok/s≈7 s32KDetails
Agentic codingwith OpenCodeComfortableMXFP4120–170 tok/s≈9 s32KDetails
Long documents (64K)with Open WebUIExcellentMXFP485–110 tok/s≈25 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.