Local AI Advisor

Can Desktop PC · GeForce RTX 5060 Ti 16GB · 32 GB RAM run Recursive Flint (40B-A2.5B)?

16 GB VRAM + 32 GB RAM · 448 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 Ollama. Generation: 60–80 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 WebUIBorderlineMXFP460–80 tok/sunder 1 s8KDetails
Coding questionswith ContinueBorderlineMXFP455–75 tok/s≈3 s16KDetails
Coding in a repositorywith AiderBorderlineMXFP445–60 tok/s≈9 s32KDetails
Agentic codingwith OpenCodeBorderlineMXFP440–55 tok/s≈15 s32KDetails
Long documents (64K)with Open WebUIBorderlineMXFP425–35 tok/s≈40 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.