DeepSeek-R1 Distill Llama 70B
DeepSeek-R1 Distill Llama 70B needs 45.8 GB of VRAM at Q4_K_M with 8k context, computed from its config.json as checked on 16 October 2026. The smallest card that holds it has 80 GB, and the fastest desktop card that runs it, the RTX PRO 6000 Blackwell, generates about 31 tokens per second.
- Parameters
- 70.6B
- Context
- 131,072 tokens
- Licence
- MIT
- Released
- January 2025
- Config checked
- 16 October 2026
Open DeepSeek-R1 Distill Llama 70B in the LLM checker
Which card size runs it
45.8 GB against each memory size. A model may use 95% of a card, and only what is left over, or missing, is coloured; tight means above 85%, with little room for a longer conversation.
Other quantisations, context lengths or a Q8 KV cache: the VRAM calculator.
Computed 16 October 2026Which GPU to pick
The smallest card here that loads it at Q4_K_M with 8k of context is the A100 (80 GB): 45.8 GB needed, an estimated 35 tokens per second. The same card still has room at 32,768 tokens of context (53.8 GB), so there is no reason to buy above it for this model.
At Q8_0, which is near-lossless, the card is the RTX PRO 6000 Blackwell (96 GB), needing 81 GB. On Apple silicon, a MacBook Pro M4 Max holds it with 32,768 tokens of context at an estimated 9 tokens per second; prompt processing is slower than on Nvidia, which long documents make obvious. Among unified-memory desktops, the Ryzen AI Max+ 395 holds it at 32,768 tokens in 53.8 GB, at an estimated 4 tokens per second.
Speed on each card
| GPU | Fits | Est. tokens/s | Feels like | Memory | Bandwidth |
|---|---|---|---|---|---|
| H100 SXM | Yes | 57 | comfortable for chat | 80 GB | 3,350 GB/s |
| A100 | Yes | 35 | comfortable for chat | 80 GB | 2,039 GB/s |
| RTX PRO 6000 Blackwell | Yes | 31 | comfortable for chat | 96 GB | 1,792 GB/s |
| RTX Spark | Yes | 5 | slow; fine for batch jobs | 128 GB (102 for the GPU) | 300 GB/s |
| DGX Spark | Yes | 5 | too slow for interactive use | 128 GB (126 for the GPU) | 273 GB/s |
| Ryzen AI Max+ 395 | Yes | 4 | too slow for interactive use | 128 GB (96 for the GPU) | 256 GB/s |
| RTX 6000 Ada | No | - | does not fit | 48 GB | 960 GB/s |
| L40S | No | - | does not fit | 48 GB | 864 GB/s |
| RTX A6000 | No | - | does not fit | 48 GB | 768 GB/s |
| RTX PRO 5000 Blackwell | No | - | does not fit | 48 GB | 1,344 GB/s |
| RTX 5090 | No | - | does not fit | 32 GB | 1,792 GB/s |
| RTX PRO 4500 Blackwell | No | - | does not fit | 32 GB | 896 GB/s |
| Radeon AI PRO R9700 | No | - | does not fit | 32 GB | 640 GB/s |
| Arc Pro B70 | No | - | does not fit | 32 GB | 608 GB/s |
| RTX 3090 | No | - | does not fit | 24 GB | 936 GB/s |
| RTX 4090 | No | - | does not fit | 24 GB | 1,008 GB/s |
| RX 7900 XTX | No | - | does not fit | 24 GB | 960 GB/s |
| RTX PRO 4000 Blackwell | No | - | does not fit | 24 GB | 672 GB/s |
| RTX PRO 4000 Blackwell SFF | No | - | does not fit | 24 GB | 432 GB/s |
| Arc Pro B60 | No | - | does not fit | 24 GB | 456 GB/s |
| RX 7900 XT | No | - | does not fit | 20 GB | 800 GB/s |
| RTX 4060 Ti | No | - | does not fit | 16 GB | 288 GB/s |
| RTX 5060 Ti | No | - | does not fit | 16 GB | 448 GB/s |
| RTX 5080 | No | - | does not fit | 16 GB | 960 GB/s |
| RTX 5070 Ti | No | - | does not fit | 16 GB | 896 GB/s |
| RTX 4080 SUPER | No | - | does not fit | 16 GB | 736 GB/s |
| RX 9070 XT | No | - | does not fit | 16 GB | 640 GB/s |
| RX 9060 XT | No | - | does not fit | 16 GB | 320 GB/s |
| RTX 3060 | No | - | does not fit | 12 GB | 360 GB/s |
| RTX 4070 | No | - | does not fit | 12 GB | 504 GB/s |
| RTX 5070 | No | - | does not fit | 12 GB | 672 GB/s |
| Arc B580 | No | - | does not fit | 12 GB | 456 GB/s |
Single-stream decode, from memory bandwidth at 70% efficiency. Prompt processing and batching not included. These are ceilings, not measurements. DeepSeek-R1 Distill Llama 70B reads 40.9 GB of weights per token at Q4, enough that the weight reads are most of the work, which is the case the formula is built for. See the speed estimator for other quantisations and Apple hardware, or every card compared if you are choosing hardware rather than a model.
Computed 16 October 2026Memory by quantisation and context
| Quant | 4,096 ctx | 8,192 ctx | 32,768 ctx |
|---|---|---|---|
| Q8_0 | 79.7 GB · fits 96 GB | 81 GB · fits 96 GB | 89.1 GB · fits 96 GB |
| Q5_K_M | 54 GB · fits 80 GB | 55.3 GB · fits 80 GB | 63.4 GB · fits 80 GB |
| Q4_K_M | 44.4 GB · fits 48 GB | 45.8 GB · fits 80 GB | 53.8 GB · fits 80 GB |
Weights at this quant: 40.9 GB at Q4. Every extra 1,000 tokens of context adds about 0.328 GB of KV cache at FP16. Try other settings in the VRAM calculator.
About the model
DeepSeek-R1 Distill Llama 70B has 70.6B parameters, 80 layers and 8 KV heads of dimension 128. DeepSeek-R1's reasoning distilled into Llama 3.3 70B, January 2025. The largest of the R1 distillations.
Text only 131,072-token window MIT January 2025
Reasoning. Always reasons, at length, before it answers.
This is a reasoning model: it writes out its thinking before it answers, often thousands of tokens of it, and that thinking occupies the context window like anything else. Budget context for the reasoning, not only for the conversation.
Native context window: 131,072 tokens. Licence: MIT, a permissive licence. Weights published in January 2025.
Good at
- Hard reasoning problems on an 80 GB card, with the working shown.
Watch out for
- Same memory as Llama 3.3 70B, but it needs much more context per answer, because the reasoning is written into the conversation.
- DeepSeek releases it under MIT, but it is built on Llama 3.3, so Meta's licence terms follow it.
- The thinking cannot be switched off.
Similar models
The models that need about the same memory as DeepSeek-R1 Distill Llama 70B at Q4_K_M and 8k, which makes them the real alternatives on whatever card you have. The context column is what each extra 1,000 tokens costs, and it is where models of the same size differ most.
| Model | Size | Needs | Per 1k context | Window | Released |
|---|---|---|---|---|---|
| DeepSeek-R1 Distill Llama 70B | 70.6B | 45.8 GB | 0.328 GB | 128k | 2025-01 |
| Llama 3.3 70B | 70.6B | 45.8 GB | 0.328 GB | 128k | 2024-12 |
| Nemotron 3.5 Bolt 74B-A9B (MoE) | 74.9B · 9B active | 45.9 GB | 0.008 GB | 256k | 2026-10 |
| Qwen3-Coder-Next 80B-A3B (MoE) | 79.7B · 3B active | 48.9 GB | 0.025 GB | 256k | 2026-01 |
| gpt-oss 120B | 117B · 5.1B active | 71.4 GB | 0.037 GB | 128k | 2025-08 |
See every model compared, which models fit on 8 to 96 GB, or the cheapest hour of every cloud card.
Questions
- Can I run DeepSeek-R1 Distill Llama 70B on a 24 GB card?
- Not comfortably. At Q4_K_M and 8k context it needs about 45.8 GB. The smallest tier that fits is 80 GB.
- How much VRAM does DeepSeek-R1 Distill Llama 70B need at Q8?
- About 81 GB at 8k context, or 89.1 GB at 32k. Q8 is near-lossless; use it when it fits.
- What is the context length of DeepSeek-R1 Distill Llama 70B?
- 131,072 tokens natively. Holding all of it at Q4_K_M takes about 86 GB, of which 42.9 GB is context.
- How fast is DeepSeek-R1 Distill Llama 70B on an RTX 4090?
- It does not fit a 4090 at Q4 and 8k context, so speed would collapse to CPU offloading. Use a larger card or a smaller quantisation.
- Architecture values from config.json in deepseek-ai/DeepSeek-R1-Distill-Llama-70B on Hugging Face, checked 16 October 2026. Every figure on this page is computed from those values; check them before buying hardware for this model.
- Estimates assume a single conversation on a card that is otherwise free. A desktop on the same GPU takes 0.5 to 2 GB.