Smaller alternatives
Models that need less memory than Llama 3.3 70B and score about as well or better.
Llama 3.3 70B Instruct
Meta · 70.6B · Dense
Llama 3.3 70B Instruct is a 70.6B model from Meta with a 128K-token context. At Q4_K_M with an 8K context it needs about 42.4 GB; 55 GB leaves room for longer chats.
Quantization options
| Quant | Bits | Memory | Quality | On your hardware |
|---|---|---|---|---|
| Q2_K | 3.16 | — | Noticeably worse | … |
| Q3_K_M | 4 | — | Some loss | … |
| Q4_K_M | 4.89 | — | Good | … |
| Q5_K_M | 5.7 | — | Good | … |
| Q6_K | 6.56 | — | Near original | … |
| Q8_0 | 8.5 | — | Near original | … |
| F16 | 16 | — | Original | … |
Memory by context length
At Q4_K_M. The context cache grows with every token the model keeps in mind.
| Context | Context cache | Total |
|---|---|---|
| 4K | 1.3 GB | 41.2 GB |
| 8K | 2.5 GB | 42.4 GB |
| 32K | 10 GB | 49.9 GB |
| 128K | 40 GB | 79.9 GB |
Next steps
Compare it with
Scored models of a similar size, side by side on your device.
Least hardware that runs it well
Desktop devices with the least memory that give Llama 3.3 70B grade A or S at Q4_K_M.
Can I run Llama 3.3 70B Instruct locally?
Can I run Llama 3.3 70B Instruct locally?
Yes, if your GPU or Mac has about 42.4 GB free for it at Q4_K_M. Open this page on that computer to see the grade for your exact hardware, then start it with llama.cpp, Ollama or LM Studio.
How much memory does Llama 3.3 70B Instruct need?
About 42.4 GB at Q4_K_M with an 8K context and 72.6 GB at Q8_0. Lower quantizations fit smaller cards with some loss in quality; longer contexts add to the total.