Qwen2.5-Coder on NVIDIA GeForce RTX 5080 · 16GB VRAM

✅ Runs (tight) — Qwen2.5-Coder 14B @ Q5_K_M
Qwen2.5-Coder 14B at Q5_K_M needs ~11.8 GB (weights 9.8 GB + KV 1.5 GB + overhead 512 MB @ 8K ctx) of your 14.4 GB usable VRAM — fits, but little headroom — close other apps or trim context. Expect ~35–55 tok/s (fast).
context @ Q5_K_M: comfortable to 4K · runs to 64K · won't fit past that on this rig
quantneedsspeed
~ Q8_016.8 GB~7–12 tok/s (usable)
✓ Q5_K_M11.8 GB~35–55 tok/s (fast)
✓ Q4_K_M10.3 GB~40–65 tok/s (very fast)
✓ Q3_K_M8.7 GB~45–80 tok/s (very fast)
$runlocal install qwen2.5-coder:14b
⚠ NVIDIA support is best-effort in v0.1 — verify before relying on it.
runlocal verdict card
download card share on x share on reddit drop it in your model-drop post
Qwen2.5-Coder: Runs (tight)
for your own README — links back here

Same model, other machines: Apple M1 · Apple M1 Pro · Apple M2 · Apple M3 Pro · Apple M2 Pro · Apple M4 Pro

Also runs on a NVIDIA GeForce RTX 5080: DeepSeek-R1 (distill) · Qwen2.5 · Qwen3 · Gemma 4 · Gemma 3

Best models for a NVIDIA GeForce RTX 5080 · 16GB VRAM →

Architecturally similar

These share Qwen2.5-Coder's KV-cache geometry — they size memory the same way, so here's how they fit on a NVIDIA GeForce RTX 5080 · 16GB VRAM too:

modelsharesverdict here
Llama 3.1 8kv/128hd ✅ Runs comfortably
Llama 3.2 8kv/128hd ✅ Runs comfortably
Llama 3.3 8kv/128hd ❌ Won't fit
Mistral 7B 8kv/128hd ✅ Runs comfortably
Mistral Small 3 8kv/128hd ✅ Runs (tight)
Mistral Nemo 8kv/128hd ✅ Runs (tight)
Codestral 8kv/128hd ✅ Runs (tight)
Llama 3.2 Vision 8kv/128hd ✅ Runs (tight)

…and 12 more — see the Qwen2.5-Coder page.

Check any combo yourself: open the checker →