DeepSeek-V3-0324

Updated

Reasoning

Code

Multilingual

Tools

Run

A 671B-parameter Mixture-of-Experts LLM from DeepSeek-AI (37B active) with a 128K context, strong coding and improved function calling.

pip install -U transformers
huggingface-cli download deepseek-ai/DeepSeek-V3-0324

curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "deepseek-ai/DeepSeek-V3-0324", "messages": [{"role": "user", "content": "Hello"}]}'

from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("deepseek-ai/DeepSeek-V3-0324", trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained("deepseek-ai/DeepSeek-V3-0324", trust_remote_code=True)

import OpenAI from "openai";
const client = new OpenAI({ baseURL: "http://localhost:8000/v1", apiKey: "local" });
const res = await client.chat.completions.create({
  model: "deepseek-ai/DeepSeek-V3-0324",
  messages: [{ role: "user", content: "Hello" }],
});
console.log(res.choices[0].message.content);

More models

View all

Name	Size / Usage	Context	Input
DeepSeek-V4-Pro		1M	Text
DeepSeek-V4-Flash		1M	Text
DeepSeek-R1-0528	Reasoning, math, coding	128K	Text
DeepSeek-Coder-V2-Lite-Instruct	Code generation, completion	128K	Text
DeepSeek-V3.2		128K	Text
DeepSeek-R1		128K	Text

At a glance

License: MIT
Context length: 128K tokens
Architecture: 671B Mixture-of-Experts, ~37B active per token
Minimum hardware: ~700 GB VRAM at FP8; ~130 GB with 1.58-bit quantization
Strengths: coding, reasoning, function calling, long-context chat

Overview

DeepSeek-V3-0324 is an open-weight large language model released by DeepSeek-AI in March 2025.

License: MIT
Context length: 128K tokens
Architecture: 671B Mixture-of-Experts, ~37B active per token
Minimum hardware: ~700 GB VRAM at FP8; ~130 GB with 1.58-bit quantization
Strengths: coding, reasoning, function calling, long-context chat

Highlights

MMLU-Pro — improves from 75.9 to 81.2 (+5.3) over the prior DeepSeek-V3.
GPQA — rises from 59.1 to 68.4 (+9.3).
AIME — jumps from 39.6 to 59.4 (+19.8).
LiveCodeBench — climbs from 39.2 to 49.2 (+10.0), with better front-end coding and function calling.

What it's good at

Compared with the first DeepSeek-V3 release, this checkpoint posts higher scores on knowledge and reasoning benchmarks, with MMLU-Pro moving from 75.9 to 81.2 and GPQA from 59.1 to 68.4. DeepSeek also reports better front-end and general code generation, steadier multi-turn rewriting, and more dependable function calling, which makes it a practical backbone for coding assistants and tool-using agents. Its 128K token context window handles long documents and large codebases in a single request. It is a general instruction and chat model, not a dedicated chain-of-thought model like DeepSeek-R1.

Running locally

The native weights ship in FP8 and occupy around 700 GB, so full-precision inference needs a multi-GPU server with comparable VRAM. Community quantizations narrow this gap: 4-bit GGUF builds land near 400 GB, and aggressive 1.58-bit dynamic quants from Unsloth drop to roughly 130 GB, runnable across high-RAM workstations or smaller clusters at lower throughput. Common serving paths include vLLM and SGLang for GPU deployment and llama.cpp for quantized CPU or mixed setups.

License

The repository and the model weights are released under the MIT License. That permits commercial use, modification, and redistribution with minimal conditions, so teams can self-host or fine-tune the model without separate licensing terms.

Desktop

macOS

(M1 or better)

Download

Windows

(x64)

Download

Linux

(x86_64)

Download

Frequently asked questions

DeepSeek-V3-0324 is an open-weight large language model released in March 2025 by DeepSeek-AI. It is a Mixture-of-Experts (MoE) model with 671B total parameters, of which about 37B are activated per token. The 0324 checkpoint is a post-training update of DeepSeek-V3 that improves reasoning, coding, and function calling over the original release.

The full FP8 weights are around 700 GB, so running the model unquantized requires a multi-GPU server with roughly 700 GB or more of combined VRAM (for example 8 to 16 high-memory data-center GPUs). Community 4-bit and 1.58-bit GGUF quantizations from projects like Unsloth bring the footprint down to roughly 130 to 400 GB, which can run across high-RAM workstations or smaller GPU clusters at reduced speed. This is not a model that fits on a single consumer card.

Yes. The repository and the model weights are released under the MIT License, which permits commercial use, modification, and redistribution. You can download the weights from Hugging Face at no cost, or access the model through DeepSeek's API and third-party inference providers, which charge per token.

DeepSeek-V3-0324 supports a 128K token context window. That capacity lets it work over long documents, large codebases, and extended multi-turn conversations within a single request.

DeepSeek-V3-0324 is a refreshed post-training checkpoint built on the same architecture as DeepSeek-V3. DeepSeek reports clear benchmark gains, including MMLU-Pro rising from 75.9 to 81.2 and GPQA from 59.1 to 68.4, along with stronger front-end code generation, more reliable multi-turn rewriting, and improved function calling. It is not a separate reasoning model like DeepSeek-R1; it remains a general-purpose chat and instruction model.