DeepSeek V4 Flash 0731 Reap 150B — Hardware Requirements & GPU Compatibility
Chat
DeepSeek V4 Flash 0731 Reap 150B is a 150.1B-parameter open language model from puwaer in the DeepSeek V4 family. It supports a context window of up to 1,048,576 tokens. At Q4_K_M it needs about 90.40 GB of VRAM — see which GPUs and Macs can run it below.
DeepSeek V4 Flash 0731 Reap 150B runs on a 96 GB+ Mac or a workstation GPU
Best local pick → Mac Studio (M3 Ultra, 96GB) · Q4_K_M · 90.6 GB · ~51 tok/s
Macs from 64 GB (IQ3_XXS · ~63 tok/s)
MIT
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run DeepSeek V4 Flash 0731 Reap 150B
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
ollama run hf.co/puwaer/DeepSeek-V4-Flash-0731-reap-150b-ggufQ4_K_M needs offload on the RTX 4090 · 66.6 GB short · 16K context
Download: DeepSeek-V4-Flash-0731-reap-150b-Q4_K_M.gguf90.08 GBpuwaer
Other ways
each opens step-by-stepRun in cloud
Fits on H200 (141 GB) (50 GB headroom) · Q4_K_M
- Generation speed
- ~227 tok/s
- generation speed
- Cost per 1M output tokens
- $4.40
- per 1M output tokens
How Much VRAM Does DeepSeek V4 Flash 0731 Reap 150B Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 58.5 GB | 58.8 GB | 3-bit · i-quant — Importance-weighted 3-bit | Does not fit: needs 58.7 GB | |
| 64.1 GB | 64.5 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Does not fit: needs 64.3 GB | |
| 73.5 GB | 73.9 GB | 3-bit medium — 3-bit medium quantization | Does not fit: needs 73.7 GB | |
| 90.4 GB | 90.7 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Does not fit: needs 90.6 GB | |
| 107.3 GB | 107.6 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Does not fit: needs 107.5 GB | |
| 124.2 GB | 124.5 GB | 6-bit, very good — 6-bit quantization, very good quality | Does not fit: needs 124.3 GB | |
| 150.4 GB | 150.8 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Does not fit: needs 150.6 GB | |
| 300.6 GB | 300.9 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Does not fit: needs 300.7 GB |
Selected Q4_K_M
4-bit medium quantization90.6GBat 16K context
- Weights
- 90.1 GB
- KV cache
- 0.2 GB
- Overhead
- 0.3 GB
Doesn't fit RTX 4090 (reference) · needs 66.6 GB more
Which Hardware Can Run DeepSeek V4 Flash 0731 Reap 150B?
Q4_K_M · 90.4 GBDeepSeek V4 Flash 0731 Reap 150B (Q4_K_M) requires 90.4 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 118+ GB is recommended. Using the full 1049K context window can add up to 11.5 GB, bringing total usage to 101.9 GB. No consumer GPU has enough memory. 30 devices with unified memory can run DeepSeek V4 Flash 0731 Reap 150B, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio (M3 Ultra, 96GB).
64–80 GB
Q4_K_M · best fit IQ3_XXS (58.7 GB)
3 Mac classes · 1 device family
M2 Ultra*64 GB~63Simulate output on Mac Pro M2 Ultra (64 GB)M5 Max64 GB~55Simulate output on Mac Studio M5 Max (64 GB)Orin~30Simulate output on NVIDIA Jetson AGX Orin 64GB96 GB
fits · 5.4 GB free · up to 256K ctx
2 Mac classes
M3 Ultra*96 GB~51Simulate output on Mac Studio (M3 Ultra, 96GB)M3 Max*96 GB~26Simulate output on MacBook Pro 14" M3 Max (96 GB)
By VRAM tier · typical GPUs and Macs at Q4_K_M
Alternatives to DeepSeek V4 Flash 0731 Reap 150B
DeepSeek V4 family · 12 models →Similar models
2| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Mixtral 8x22B Instruct v0.1Mistral AI | 71.6 | 88.4 GB | —needs > 24 GB | same size |
| WizardLM 2 8x22Balpindale | — | 88.4 GB | —needs > 24 GB | same size |
Specifications
- Publisher
- puwaer
- Family
- DeepSeek V4
- Parameters
- 150.1B
- Active parameters
- 10.6B per token
- Architecture
- DeepseekV4ForCausalLM
- Context Length
- 1,048,576 tokens
- Vocabulary Size
- 129,280
- Release Date
- 2026-08-09
- License
- MIT
Frequently Asked Questions
How much VRAM does DeepSeek V4 Flash 0731 Reap 150B need?
DeepSeek V4 Flash 0731 Reap 150B requires 90.4 GB of VRAM at Q4_K_M, or 300.6 GB at BF16. Full 1049K context adds up to 11.5 GB (101.9 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 150.1B × 4.8 bits ÷ 8 = 90.1 GB
KV Cache + Overhead ≈ 0.3 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 ≈ 11.8 GB (at full 1049K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run DeepSeek V4 Flash 0731 Reap 150B?
No — DeepSeek V4 Flash 0731 Reap 150B requires at least 58.5 GB at IQ3_XXS, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
What's the best quantization for DeepSeek V4 Flash 0731 Reap 150B?
For DeepSeek V4 Flash 0731 Reap 150B, Q4_K_M (90.4 GB) offers the best balance of quality and VRAM usage. Q5_K_M (107.3 GB) provides better quality if you have the VRAM. The smallest option is IQ3_XXS at 58.5 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run DeepSeek V4 Flash 0731 Reap 150B on a Mac?
Yes — Mac Pro M2 Ultra (64 GB) and 41 other Macs can run DeepSeek V4 Flash 0731 Reap 150B, the smallest at IQ3_XXS. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 90.4 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio (M3 Ultra, 96GB).
Can I run DeepSeek V4 Flash 0731 Reap 150B locally?
Yes — DeepSeek V4 Flash 0731 Reap 150B can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is DeepSeek V4 Flash 0731 Reap 150B?
At Q4_K_M, DeepSeek V4 Flash 0731 Reap 150B can reach ~101 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 → 6.4 GB active ÷ (8000 × 0.65) = 1.23 ms, plus 43 layers × 0.055 ms = 2.37 ms, so 1000 ÷ 3.60 ms = ~278 tok/s
Estimated speed at Q4_K_M (90.4 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 Reap 150B?
At Q4_K_M, the download is about 90.08 GB. The full-precision BF16 version is 300.26 GB. The smallest option (IQ3_XXS) is 58.17 GB.
Which GPUs can run DeepSeek V4 Flash 0731 Reap 150B?
No single consumer GPU has enough VRAM to run DeepSeek V4 Flash 0731 Reap 150B at Q4_K_M (90.4 GB). Multi-GPU or professional hardware is required.
Which devices can run DeepSeek V4 Flash 0731 Reap 150B?
30 devices with unified memory can run DeepSeek V4 Flash 0731 Reap 150B at Q4_K_M (90.4 GB), including ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB), GMKtec EVO-X2 (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.