Keimodel
Log inStart free
DeepSeek

DeepSeek

Chinese frontier AI at remarkably low cost

DeepSeek is a Chinese AI lab that stunned the industry with DeepSeek-R1, a reasoning model that matches OpenAI's o1 while being trained at a fraction of the cost. Their mixture-of-experts DeepSeek V3 architecture and open-weight releases make them a disruptive force. DeepSeek models are available via OpenRouter at significantly lower prices than comparable Western models.

Founded 2023·6 models·Visit DeepSeek

6 Models

DeepSeek V4 Pro 0813

DeepSeek's frontier open-weights model

3 cr
DeepSeek V4 Pro 0813 is a large-scale mixture-of-experts model from DeepSeek. This is the GA release of DeepSeek V4 Pro.
1M ctxTool useJSON mode

DeepSeek V4 Flash 0731

Very cheap long-context DeepSeek

2 cr
DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total. This re-post-trained revision is suited for coding, reasoning, and agent workflows....
1M ctxTool useJSON mode

DeepSeek V3.2

The V3 generation workhorse

3 cr
DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism...
164k ctxTool useJSON mode

DeepSeek V3

685B MoE model for general intelligence

3 cr
DeepSeek V3 is a 685-billion parameter mixture-of-experts model that activates only 37B parameters per forward pass, making it highly efficient. It outperforms all previous open-source models on coding, math, and reasoning.
163k ctxTool useJSON mode

DeepSeek R1

Open-weight reasoning model matching o1

5 cr
DeepSeek R1 is a pure reasoning model trained with reinforcement learning on chain-of-thought traces. It matches OpenAI's o1 on math and coding benchmarks while being open-weight and significantly cheaper to run via API.
163k ctxJSON modeCode

DeepSeek R1 Distill 70B

Reasoning capabilities distilled into 70B Llama

3 cr
DeepSeek R1 Distill 70B transfers the chain-of-thought reasoning abilities of the full R1 model into a 70B Llama architecture via knowledge distillation. It offers strong reasoning performance at a fraction of the full model's cost.
128k ctxCode

Other providers

OpenAIOpenAIAnthropicAnthropicGoogleGoogleMetaMeta