Mistral Small 3 on NVIDIA GeForce RTX 5080 · 16GB VRAM

✅ Runs (tight) — Mistral Small 3 24B @ Q3_K_M
Mistral Small 3 24B at Q3_K_M needs ~12.5 GB (weights 10.7 GB + KV 1.3 GB + overhead 547 MB @ 8K ctx) of your 14.4 GB usable VRAM — fits, but little headroom — close other apps or trim context. Expect ~30–50 tok/s (fast).
context @ Q3_K_M: runs to its full 32K
quantneedsspeed
~ Q8_025.8 GB—
~ Q5_K_M17.6 GB~7–12 tok/s (usable)
~ Q4_K_M15.3 GB~8–13 tok/s (usable)
✓ Q3_K_M12.5 GB~30–50 tok/s (fast)
$runlocal install mistral-small:24b
⚠ 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
Mistral Small 3: 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 Mistral Small 3'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
Qwen2.5-Coder 8kv/128hd ✅ Runs (tight)
Llama 3.2 8kv/128hd ✅ Runs comfortably
Llama 3.3 8kv/128hd ❌ Won't fit
Mistral 7B 8kv/128hd ✅ Runs comfortably
Mistral Nemo 8kv/128hd ✅ Runs (tight)
Codestral 8kv/128hd ✅ Runs (tight)
Llama 3.2 Vision 8kv/128hd ✅ Runs (tight)

…and 8 more — see the Mistral Small 3 page.

Check any combo yourself: open the checker →