OpenModelsHQ

DeepSeek-V3

DeepSeekLLMMoENot statedChina Desk

DeepSeek-V3 is an open-weight language model from DeepSeek with 684B parameters (mixture of experts). At 4-bit it needs about 467.5 GB of memory, more than any single device in our directory has.

Overview

Estimated memory at 4-bit
467.5GB

Includes room for an 8k context and runtime overhead. More context needs more memory.

Total parameters
684B
Mixture of experts, all weights counted
Context window
160K
163,840 tokens
Cheapest API, blended / 1M
$0.37
₹35 · DeepInfra · 3:1 input to output
All DeepSeek API prices

Key information

Lab
DeepSeek
Released
25 Dec 2024
Parameters
684B · MoE
Type
Language
Context
163,840 tokens
License
Not stated
HF downloads, 30 days
1,413,463
View all

Will it run on your hardware?

How we estimate
Memory needed vs devices
DeepSeek-V3 needs about 467.5 GB at 4-bit, 813.9 GB at 8-bit and 1521.6 GB at 16-bit. At 4-bit it fits none of the reference devices. Dashed lines are usable memory on each device; a model fits when its estimate stays within 90% of it. Log scale. 8k context.
Memory needed vs devices: data
Estimate or deviceMemory
Q4467.5 GB
Q8813.9 GB
BF161521.6 GB
RTX 5060 Ti 16 GB16 GB usable
RTX 5090 32 GB32 GB usable
DGX Spark 128 GB108.8 GB usable
Mac Studio M5 Ultra 256 GB192 GB usable

Single-device memory estimate. Pick a device to update the On your device column.

36 GB usable memory

FormatMemory neededOn your deviceExample hardware
4-bit (Q4)≈ 467.5 GBToo big 1299%No single device in our directory has enough memory
8-bit (Q8)≈ 813.9 GBToo big 2261%No single device in our directory has enough memory
16-bit (BF16)≈ 1521.6 GBToo big 4227%No single device in our directory has enough memory

Fits means the estimate stays within 90% of usable memory. Tight is above 90% and up to 100%. Too big is above 100%. These are memory checks, not speed or compatibility tests.

Mac allowances are 66% of total RAM up to 36 GB, and 75% above. AI boxes allow 85% because their memory is shared with the operating system.

Where to run it

On a Mac

No trusted MLX build found yet.

Check your Mac

With GGUF

Choose a file for llama.cpp or a compatible app.

unsloth/DeepSeek-V3-GGUFFind compatible hardware

Through an API

DeepInfra (fp4) is the cheapest provider we found.

$0.256 / $0.712

Input / output per 1M tokens
₹24 / ₹67 at ₹94.29 per $1

Blended: $0.37 (₹35) per 1M. Ranked by a 3:1 mix of input and output tokens.

GGUF with llama.cppView repository
llama-cli -hf unsloth/DeepSeek-V3-GGUF \
  -c 8192 -p "Explain how a transformer works."

Example commands built from the repository ids above. They need a working install of llama.cpp. We have not run them; check each repository for available files.

Hosted API 2

USD and INR per million tokensAll DeepSeek API prices
Precision
DeepInfraCheapest$0.37₹35$0.256₹24$0.712₹67163,840FP4
StreamLake$0.4502₹42$0.2574₹24$1.0287₹97128,000Not listed
View on OpenRouterEndpoint sourceSorted by the blended price: 3 parts input to 1 part output. Context limits differ by provider.

INR at ₹94.29 per $1. Prices fetched 14 Nov 2026.

More about this model

Frequently asked questions

How much memory does DeepSeek-V3 need? At an 8k context, about 467.5 GB at 4-bit, 813.9 GB at 8-bit and 1521.6 GB at 16-bit (BF16). A longer context needs more.

Can I run DeepSeek-V3 on a Mac? No Mac in our directory has enough usable memory for the 4-bit estimate of 467.5 GB. We have not found a trusted MLX build yet, but a GGUF build is available.

Does the full context fit? DeepSeek-V3 supports up to 163,840 tokens of context. Our memory estimates use 8k, so long prompts need more memory than shown.

What does it cost through an API? The cheapest provider we found is DeepInfra, at $0.26 per 1M input tokens and $0.71 per 1M output tokens (₹24 and ₹67 at 94.29 rupees per dollar). Ranked by a 3:1 mix of input and output tokens.

License in plain English

Not stated. No license is listed in this model's metadata. Check the model card before using it. This summary is not legal advice.

Open the model card

OMHQ RigBench

Not yet tested. We show no speed or quality score without a measurement.

Revision history and sources

Last updated . Sources:Hugging Face model card, GGUF repository, OpenRouter.