DeepSeek V3
DeepSeek V3 — DeepSeek large language model (2024) with a 128K-token context window and an MMLU score of 88.5%, released with open weights.
Key specifications
| MMLU (capability) | 88.5% |
|---|---|
| Context window (K tokens) | 128 |
| Input price ($/1M) | 0.3 |
| Output price ($/1M) | 1.1 |
About DeepSeek V3
The DeepSeek V3 is a large language model from DeepSeek with a 128K-token context window and an MMLU score of 88.5%. It is open weights, so it can be self-hosted. MMLU measures general knowledge and reasoning; real-world quality also depends on coding, tool use, latency and safety.
The comparisons below show how DeepSeek V3 stacks up against the alternatives people most often weigh against it, with the specific numbers laid out side by side.
Frequently asked questions
Is the DeepSeek V3 any good?
DeepSeek V3 — DeepSeek large language model (2024) with a 128K-token context window and an MMLU score of 88.5%, released with open weights.
What are the key specs of the DeepSeek V3?
See the specification table above for the full breakdown.