Best local LLM for 32GB
Short answer: Nemotron 3.5 Bolt 74B-A9B at IQ2_XXS for general use — 23.1 GB of 32GB, leaving 9.0 GB. Also: .
“Best” here means the largest model in its category that clears the card with headroom left, at the highest-quality quant that still fits. This index measures memory and published perplexity, not task quality — it cannot tell you which model answers better, and does not pretend to.
The picks
| For | Model | Quant | Est. VRAM | Headroom | Loss | Why |
|---|---|---|---|---|---|---|
| General · Coding · Images | Nemotron 3.5 Bolt 74B-A9B74B-A9B | IQ2_XXS | 23.1 GB | 9.0 GB | — | The largest model of any kind that fits this budget. Its quantizer published no perplexity figure, so none is quoted. It still fits at 32K context, which is what the headroom is for. |
Sized at 4K context with a 10% activation buffer; 90 of 112 models in the index clear this budget comfortably. Loss is the quantizer’s published perplexity figure against FP16, or a dash where none was published.
Cards this applies to
What fits is decided by capacity alone, so every card here returns the same list. What differs is how fast it reads the weights — each card’s page states its bandwidth and what that costs you.
What does not fit
The nearest miss is Llama 3.1 70B Instruct. Its smallest build in this index, EXL2 3.5bpw, needs about 33.6 GB — 1.6 GB over 32GB, before any context beyond 4K. Offloading to system RAM will load it at a speed set by your RAM, which is a different machine rather than a smaller quant.
Pushing context
At 4K context 90 models clear this budget. At 16K it is 90, and at 32K it is 89. The weights do not change — the KV cache is what grows.
Check your exact context lengthCommon questions
What is the best local LLM for 32 GB of VRAM in 2026?
Nemotron 3.5 Bolt 74B-A9B at IQ2_XXS for general use — 23.1 GB of 32 GB, leaving 9.0 GB. . "Best" here means the largest model in its category that clears the card with headroom — this index measures memory and published perplexity, not task quality, so it cannot rank models on how well they answer.
How many models fit 32 GB?
90 of the 112 models in this index clear it comfortably at 4K context — comfortably meaning the estimate uses at most 88% of the memory, so there is margin for the display and the runtime. At 16K that falls to 90, and at 32K to 89: the KV cache is what moves, not the weights.
What will not run on 32 GB?
The nearest miss is Llama 3.1 70B Instruct: its smallest build here, EXL2 3.5bpw, needs about 33.6 GB — 1.6 GB over, before any context beyond 4K. Offloading part of it to system RAM will load it, at a speed set by your RAM rather than your card; that is a different machine, not a smaller quant.
Do all 32 GB cards run the same models?
They hold the same models — what fits is decided by capacity alone, so all 4 cards at this tier return this same list. What differs is how fast they read the weights: token generation reads the whole weight set once per token, so memory bandwidth sets the ceiling, and two cards with identical memory can be more than 2.5× apart on it. Each card's own page states its bandwidth and what that implies.