Local AI Advisor

Can Desktop PC · Intel Arc B580 · 32 GB RAM run Recursive Flint (40B-A2.5B)?

12 GB VRAM + 32 GB RAM · 456 GB/s memory bandwidth · Recursive Flint (40B-A2.5B) has 39.8B parameters (2.5B active per token).

Short answer

It loads, but it's borderline even for chat.

Best quantization for chat: MXFP4 (20 GB of weights) on LM Studio. Generation: 40–55 tok/s, leaving about 0.0 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 WebUIBorderlineMXFP440–55 tok/s≈2 s8KDetails
Coding questionswith ContinueBorderlineMXFP435–50 tok/s≈10 s16KDetails
Coding in a repositorywith AiderBorderlineMXFP430–40 tok/s≈30 s32KDetails
Agentic codingwith OpenCodeBorderlineMXFP425–35 tok/s≈45 s32KDetails
Long documents (64K)with Open WebUIBorderlineMXFP418–25 tok/s≈2 min64KDetails

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.