Model → hardware
Phi-4-mini-instruct
Plan for 6 GB of dedicated GPU memory at Q4_K_M and a 4,000-token context. 8 GB gives more headroom under the same settings.
Small instruction-following text model. Memory fit is an estimate, not a speed or quality guarantee.
Hardware starting points
Planning estimatesQ4_K_M · 32 GB system RAM · one active request · full GPU residency
| Dedicated GPU memory | At 4,000 tokens | At 16,000 tokens | Example from these guides |
|---|---|---|---|
| 6 GB | Smallest listed tier that fits4.9 GB + 0.18 GB reserve | Fits the estimate5.4 GB + reserve | RTX 2060 6 GB |
| 8 GB | More headroom4.9 GB + 0.24 GB reserve | Fits the estimate5.4 GB + reserve | RTX 4060 8 GB |
| 12 GB | More headroom4.9 GB + 0.36 GB reserve | Fits the estimate5.4 GB + reserve | RTX 3060 12 GB |
The smallest tier fits the calculation; it is not a verified minimum. Cache precision and runtime buffers can change the fit. “More headroom” means at least 2 GB remains after the estimate and reserve.
The download is a GGUF conversion; Q5 in the table is an estimate, not a listed file in that repository. Use the publisher’s Instruct / chat template in a current GGUF-capable runtime. The 6 GB tier can become tight when display use, buffers or context grow. CPU offloading needs adequate system RAM and changes performance; it is not included in these full-GPU estimates.
Compare quantisation and context
| File profile | Weight input¹ | GPU estimate at 4K | GPU estimate at 16K |
|---|---|---|---|
| Q4_K_M | 3.80 GB | 4.9 GB | 5.4 GB |
| Q5_K_M | 4.60 GB | 5.7 GB | 6.2 GB |
¹ Catalogue planning weights. These can differ from an exact downloaded file; runtime memory also includes cache and buffers.
Keep the exact file, runtime version, cache type and context limit consistent when comparing. Q5 uses more memory; compare its answers on your own tasks before accepting the trade-off.
Setup and common questions
Which Phi model is this?
Use Microsoft’s Phi-4-mini-instruct, a 3.8B-parameter text model under MIT. Phi-4-mini-reasoning and Phi-4-multimodal are different checkpoints; these memory estimates do not cover them.
Which file should I use?
LM Studio Community provides a matching Q4_K_M GGUF conversion of Microsoft’s checkpoint, about 2.49 GB before cache and buffers. The Finder uses a conservative 3.80 GB weight input. The Q5 table is a planning profile, not a file in that repository. Keep the correct chat template.
Why start at 4K when the model supports 128K?
The model’s 128K limit describes supported sequence length, not free memory. More context increases the cache. Start small, check runtime memory and compare answers on your task before increasing it. A small model can still give incorrect facts.
How are these recommendations chosen?
We compare dedicated GPU memory tiers using the main Finder’s memory calculation at the stated quantisation and context. These tiers describe memory fit, not speed or output quality.
GPU memory includes weights, context cache and a 0.9 GB runtime allowance, plus a 3% reserve (0.15–1.5 GB). Cache format and runtime buffers can change the fit. See the calculation assumptions.
Sources and related guides
Specification sources checked 3 October 2026.
These comparisons are planning estimates. No reviewed or independently reproduced measurements are available yet. About test evidence.