Best local LLM for 8GB
Short answer: Stable LM 2 12B Chat at AWQ INT4 for general use — 7.0 GB of 8GB, leaving 1.0 GB. Also: Falcon 3 10B Instruct (Coding), Qwen3-VL 8B Instruct (Images).
“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 | Stable LM 2 12B Chat12B | AWQ INT4 | 7.0 GB | 1.0 GB | 4.5% | The largest model of any kind that fits this budget. At AWQ INT4 it gives up 4.5% perplexity against FP16. Only at 4K context — a longer window does not fit. |
| Coding | Falcon 3 10B Instruct10B | GPTQ INT4 | 6.1 GB | 1.9 GB | 4.8% | The largest in its category that still leaves headroom. At GPTQ INT4 it gives up 4.8% perplexity against FP16. Only at 4K context — a longer window does not fit. |
| Images | Qwen3-VL 8B Instruct8B | Q4_K_M | 6.2 GB | 1.8 GB | 3.0% | The largest in its category that still leaves headroom. At Q4_K_M it gives up 3.0% perplexity against FP16. Only at 4K context — a longer window does not fit. |
Sized at 4K context with a 10% activation buffer; 35 of 81 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 Phi-3 Medium 14B Instruct. Its smallest build in this index, AWQ INT4, needs about 8.2 GB — 0.2 GB over 8GB, 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 35 models clear this budget. At 16K it is 24, and at 32K it is 16. The weights do not change — the KV cache is what grows.
Check your exact context lengthCommon questions
What is the best local LLM for 8 GB of VRAM in 2026?
Stable LM 2 12B Chat at AWQ INT4 for general use — 7.0 GB of 8 GB, leaving 1.0 GB. Falcon 3 10B Instruct if you want code, Qwen3-VL 8B Instruct if you want images. "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 8 GB?
35 of the 81 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 24, and at 32K to 16: the KV cache is what moves, not the weights.
What will not run on 8 GB?
The nearest miss is Phi-3 Medium 14B Instruct: its smallest build here, AWQ INT4, needs about 8.2 GB — 0.2 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 8 GB cards run the same models?
They hold the same models — what fits is decided by capacity alone, so all 6 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.