DeepSeek-V3
671B MoEDeepSeek
DeepSeek-V3 frontier MoE (~37B active / 671B total). MLA + FP8; multi-node GPU cluster required at Q4. MLA cache: a GGUF converted by current llama.cpp (attn_k_b / attn_v_b) holds about 2 GB at 32K, which the estimates here assume; an older conversion with a single attn_kv_b tensor caches full keys and values — about 150 GB at 32K.
164K
Max Context
2
Quant Variants
GGUF Q4_K_M
Best Quality
98.0%
Accuracy Retained
Quantization Variants
Per-quant VRAM, quality loss, and inference speed on RTX 4090
Measured = run on this site · Estimated = not run here: calculated, or a figure this site has not verified · Community = public reports
Similar models
Compare with DeepSeek-R1DeepSeek-R1
DeepSeek
DeepSeek-R1 reasoning model built on V3 MoE. Chain-of-thought at frontier scale — use distill variants for local GPUs. MLA cache: a GGUF converted by current llama.cpp (attn_k_b / attn_v_b) holds about 2 GB at 32K, which the estimates here assume; an older conversion with a single attn_kv_b tensor caches full keys and values — about 150 GB at 32K.
DeepSeek-R1-Distill-Llama-70B
DeepSeek
R1 reasoning in Llama 70B architecture. Top open reasoning model for dual-GPU setups.
DeepSeek-R1-Distill-Qwen-32B
DeepSeek
R1 distilled to 32B. Near-frontier reasoning on a single 24GB card (Q3/Q4).
DeepSeek-R1-Distill-Qwen-14B
DeepSeek
R1 reasoning distilled into 14B. Huge community interest; excellent chain-of-thought.
Running DeepSeek-V3 locally
At Q4_K_M and 4K of context, DeepSeek-V3 needs about 425.4 GB — 386.4 GB of weights, 0.27 GB of KV cache and a 38.7 GB activation buffer. No card in this index clears that comfortably at Q4_K_M, so at that level it is a multi-GPU or CPU-offload proposition. These are calculated figures, not measurements: the estimate stops counting a card as comfortable at 88% of its VRAM, which is roughly the room a desktop session needs.
What longer context costs
Going from 4K to 32K adds about 1.87 GB, taking the total to 427.4 GB. Weights do not move with context — only the KV cache does, and it grows linearly, so this is the number to watch when planning for long documents. The model's native window is 160K; holding all of it at Q4_K_M would need about 437 GB.
Which build to download
This index only tracks DeepSeek-V3 in GGUF, across 2 levels (Q4_K_M, Q3_K_M). Q4_K_M carries the lowest published perplexity loss at 2.0%. No throughput has been measured for this model on this site's hardware. GGUF runs on llama.cpp and Ollama across NVIDIA, AMD and Apple silicon; AWQ and GPTQ target vLLM on CUDA and ROCm; EXL2 is ExLlamaV2 — CUDA only, and now archived, so local use rather than serving.
Common questions
- How much VRAM does DeepSeek-V3 need?
- About 425.4 GB at Q4_K_M with 4K of context and batch 1, rising to roughly 427.4 GB at 32K. That figure is weights plus KV cache plus a 10% activation buffer, calculated from the model's architecture rather than measured on a card.
- Can DeepSeek-V3 run on a single GPU?
- Not comfortably on any single card in this index at Q4_K_M. It needs about 425.4 GB, which means splitting across GPUs or offloading layers to system RAM.
- Which quantization of DeepSeek-V3 should I use?
- Q4_K_M has the lowest published quality loss (2.0%), and Q4_K_M is the level most people run. All 2 levels in the index are Q4_K_M, Q3_K_M. Note this is a mixture-of-experts model — all parameters must be resident even though only a fraction are active per token, so the memory cost follows the total, not the active count.
Where this model fits
Sized at Q4_K_M with a 4K context window, smallest card first. Comfortable means the estimate uses at most 88% of the memory.
- Ships in
- GGUF
- Try it yourself
- Size it in the calculatorGenerate the run command
How to actually run this
Deployment guides for this model and this class of hardware.
This page's figures change when the model or the runtime does.Last updated 2026-10-20 RSS → /feed.xml