DeepSeek V4 Flash 0731 Abliterated — Hardware Requirements & GPU Compatibility
Chat
DeepSeek V4 Flash 0731 Abliterated is a 305.7B-parameter open language model from cebeuq in the DeepSeek V4 family. It supports a context window of up to 1,048,576 tokens. At Q4_K_M it needs about 183.76 GB of VRAM — see which GPUs and Macs can run it below.
DeepSeek V4 Flash 0731 Abliterated runs on a 192 GB+ Mac or a datacenter GPU
Best local pick → Mac Pro M2 Ultra (192 GB) · Q4_K_M · 183.9 GB · ~25 tok/s
Macs from 192 GB (Q4_K_M · ~25 tok/s)
MIT
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run DeepSeek V4 Flash 0731 Abliterated
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
vllm serve cebeuq/DeepSeek-V4-Flash-0731-abliterated --max-model-len 16384 --tensor-parallel-size 8vLLM repo with its own quantisation
Download: Weights on Hugging Face
Other ways
each opens step-by-stepRun in cloud
Fits on B300 (288 GB) (104 GB headroom) · Q4_K_M
- Generation speed
- ~175 tok/s
- generation speed
- Cost per 1M output tokens
- $11.00
- per 1M output tokens
How Much VRAM Does DeepSeek V4 Flash 0731 Abliterated Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | Runs on | |
|---|---|---|---|---|
| 130.3 GB | 130.6 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Runs on H200 NVL 141 GB | |
| 149.4 GB | 149.7 GB | 3-bit medium — 3-bit medium quantization | Runs on B200 192 GB | |
| 183.8 GB | 184.1 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Runs on B200 192 GB | |
| 218.2 GB | 218.5 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Runs on Instinct MI325X 256 GB | |
| 252.5 GB | 252.9 GB | 6-bit, very good — 6-bit quantization, very good quality | Runs on Instinct MI325X 256 GB | |
| 306.1 GB | 306.4 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Runs on 5× H100 SXM 80 GB | |
| 611.8 GB | 612.1 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Runs on 5× H200 SXM 141 GB |
Selected Q4_K_M
4-bit medium quantization183.9GBat 16K context
- Weights
- 183.4 GB
- KV cache
- 0.2 GB
- Overhead
- 0.3 GB
Server-class: ~183.9 GB at 16K context · one B200 192 GB
Which Hardware Can Run DeepSeek V4 Flash 0731 Abliterated?
Q4_K_M · 183.8 GBDeepSeek V4 Flash 0731 Abliterated (Q4_K_M) requires 183.8 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 239+ GB is recommended. Using the full 1049K context window can add up to 11.5 GB, bringing total usage to 195.3 GB. No consumer GPU has enough memory. 4 devices with unified memory can run DeepSeek V4 Flash 0731 Abliterated, including Mac Studio (M3 Ultra, 512GB), Mac Studio (M3 Ultra, 256GB).
192+ GB
fits · 8.1 GB free · up to 512K ctx
6 GPUs · 1 Mac class · 2 device families
B200~175Simulate output on B200B300~175Simulate output on B300Instinct MI350X~82Simulate output on Instinct MI350XM3 Ultra256–512 GB~26Simulate output on Mac Studio (M3 Ultra, 256GB)8x H100 SXM5~297Simulate output on NVIDIA DGX H1008x A100 SXM4~250Simulate output on NVIDIA DGX A100 640GB
By VRAM tier · typical GPUs and Macs at Q4_K_M
Consumer GPUs (up to 32 GB) and Macs up to 128 GB: too small for this model (needs ≥ 130.4 GB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →
Alternatives to DeepSeek V4 Flash 0731 Abliterated
DeepSeek V4 family · 12 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| GLM 5.3 FlashZ.ai | 124.2 | 205.2 GB | —needs > 24 GB | same size |
| IQuest Q1IQuestLab | — | 195.4 GB | —needs > 24 GB | same size |
| MiMo V2.6 Flash MOPDXiaomi | — | 186.5 GB | —needs > 24 GB | same size |
| Hy3Tencent | — | 182.3 GB | —needs > 24 GB | same size |
| Step3StepFun | — | 219.1 GB | —needs > 24 GB | same size |
| Motif 3Motif-Technologies | — | 632.8 GB | —needs > 24 GB | same size |
Specifications
- Publisher
- cebeuq
- Family
- DeepSeek V4
- Parameters
- 305.7B
- Active parameters
- 28.9B per token
- Architecture
- DeepseekV4ForCausalLM
- Context Length
- 1,048,576 tokens
- Vocabulary Size
- 129,280
- Release Date
- 2026-08-01
- License
- MIT
Frequently Asked Questions
How much VRAM does DeepSeek V4 Flash 0731 Abliterated need?
DeepSeek V4 Flash 0731 Abliterated requires 183.8 GB of VRAM at Q4_K_M, or 611.8 GB at BF16. Full 1049K context adds up to 11.5 GB (195.3 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 305.7B × 4.8 bits ÷ 8 = 183.4 GB
KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 11.9 GB (at full 1049K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run DeepSeek V4 Flash 0731 Abliterated?
No — DeepSeek V4 Flash 0731 Abliterated requires at least 130.3 GB at Q2_K, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
What's the best quantization for DeepSeek V4 Flash 0731 Abliterated?
For DeepSeek V4 Flash 0731 Abliterated, Q4_K_M (183.8 GB) offers the best balance of quality and VRAM usage. Q5_K_M (218.2 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 130.3 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run DeepSeek V4 Flash 0731 Abliterated on a Mac?
Yes — Mac Pro M2 Ultra (192 GB) and 3 other Macs can run DeepSeek V4 Flash 0731 Abliterated. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 183.8 GB of usable unified memory (RAM minus macOS overhead).
Can I run DeepSeek V4 Flash 0731 Abliterated locally?
Yes — DeepSeek V4 Flash 0731 Abliterated can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is DeepSeek V4 Flash 0731 Abliterated?
At Q4_K_M, DeepSeek V4 Flash 0731 Abliterated can reach ~82 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)
Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.
Example: NVIDIA B200 → 17.4 GB active ÷ (8000 × 0.65) = 3.34 ms, plus 43 layers × 0.055 ms = 2.37 ms, so 1000 ÷ 5.71 ms = ~175 tok/s
Estimated speed at Q4_K_M (183.8 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of DeepSeek V4 Flash 0731 Abliterated?
At Q4_K_M, the download is about 183.43 GB. The full-precision BF16 version is 611.45 GB. The smallest option (Q2_K) is 129.93 GB.
Which GPUs can run DeepSeek V4 Flash 0731 Abliterated?
No single consumer GPU has enough VRAM to run DeepSeek V4 Flash 0731 Abliterated at Q4_K_M (183.8 GB). Multi-GPU or professional hardware is required.
Which devices can run DeepSeek V4 Flash 0731 Abliterated?
4 devices with unified memory can run DeepSeek V4 Flash 0731 Abliterated at Q4_K_M (183.8 GB), including Mac Pro M2 Ultra (192 GB), Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio M2 Ultra (192 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.