Local AI Advisor

Can Desktop PC · GeForce RTX 5090 · 64 GB RAM run Recursive Flint (40B-A2.5B)?

32 GB VRAM + 64 GB RAM · 1792 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 Ollama. Generation: 500–680 tok/s, leaving about 10 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 WebUIExcellentMXFP4500–680 tok/sunder 1 s8KDetails
Coding questionswith ContinueExcellentMXFP4470–640 tok/sunder 1 s16KDetails
Coding in a repositorywith AiderComfortableMXFP4390–530 tok/s≈1 s32KDetails
Agentic codingwith OpenCodeComfortableMXFP4370–490 tok/s≈1 s32KDetails
Long documents (64K)with Open WebUIExcellentMXFP4250–340 tok/s≈4 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.