# DeepSeek V3

DeepSeek V3 is an AI model released by DeepSeek on Dec 26 2024. It has 671B parameters, a 128k context window and open weight. Tracked results include 59.1% on GPQA Diamond, 42% on SWE-Bench Verified and 40.5% on LiveCodeBench.

## Facts

| Field | Value |
| --- | --- |
| Model | DeepSeek V3 |
| Developer | DeepSeek |
| Release date | Thursday, Dec 26 2024 |
| Licensing | Open Weight |
| Parameters | 671B |
| Context window | 128k |

## Tracked benchmark scores

| Benchmark | Score | Source | What it measures |
| --- | --- | --- | --- |
| SWE-Bench Verified | 42% | Lab | Real coding tasks pulled from open-source projects — the AI has to find and fix actual bugs. A human-checked version of the original SWE-Bench. Higher is better. |
| LiveCodeBench | 40.5% | Lab | Coding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better. |
| GPQA Diamond | 59.1% | Lab | Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. |
| MMLU | 88.5% | Lab | A 57-subject multiple-choice exam — history, law, medicine, maths — that was the standard measure of how much a model knows from 2020 until roughly 2024, when frontier scores crowded into the high 80s and labs moved on to harder tests. The scores here were published years apart under different testing setups, so read them as a historical record rather than a like-for-like ranking. Higher is better. |

## About DeepSeek V3

DeepSeek V3 was released on December 26, 2024 as a mixture-of-experts language-model family with 671 billion main-model parameters and about 37 billion active per token. DeepSeek published the pretrained Base checkpoint and the post-trained chat model together. Base offered a foundation for further adaptation; the chat model added supervised fine-tuning and reinforcement learning for assistant use. All benchmark scores displayed here represent the chat model.

V3 was pretrained on 14.8 trillion tokens. It retained Multi-head Latent Attention from V2 and introduced auxiliary-loss-free expert balancing, FP8 mixed-precision training and a multi-token prediction objective to improve efficiency. The release package included additional prediction-module weights beyond the headline main-model size. This entry groups the simultaneous Base and chat release while keeping their evaluation results separate.

## Questions and answers

### When was DeepSeek V3 released?

DeepSeek V3 was released by DeepSeek on Thursday, Dec 26 2024.

### Who made DeepSeek V3?

DeepSeek V3 was built by DeepSeek. Chinese AI lab known for efficient, open-weight models. Gained attention for strong performance at lower cost.

### What benchmark scores did DeepSeek V3 get?

DeepSeek V3 reports 4 tracked benchmark scores — SWE-Bench Verified: 42%; LiveCodeBench: 40.5%; GPQA Diamond: 59.1%; MMLU: 88.5%. Tracked scores may come from lab reports or independent benchmarks; source details accompany the benchmark data.

### What is the context window of DeepSeek V3?

DeepSeek V3 has a context window of 128k. Providers may specify separate input and output token limits.

### How many parameters does DeepSeek V3 have?

DeepSeek V3 is reported at 671B parameters.

### Is DeepSeek V3 open source?

Partly. DeepSeek V3 is an open-weight model: the trained weights are free to download and run locally or on your own infrastructure, but the training data and code are not fully released and the license may restrict some commercial uses. It is not open source in the strict sense.

### What came before and after DeepSeek V3?

DeepSeek's previous tracked release was DeepSeek V2.5 (Revised) on Dec 10 2024, 16 days earlier. It was followed by DeepSeek R1 on Jan 20 2025.


---

Canonical page: https://aireleasetracker.com/model/deepseek/deepseek-v3
Site index: https://aireleasetracker.com/llms.txt
Source: AI Release Tracker (https://aireleasetracker.com). Most benchmark scores come from lab launch material; gathered results identify the leaderboard that published them.
