DeepSeek Pay Per Use
Access reviewed DeepSeek chat and code models with token-metered pricing through an OpenAI-compatible API.
Why DeepSeek is gaining traction
DeepSeek provides chat, reasoning, and code-completion models through an OpenAI-compatible API. Locus exposes a reviewed set of those operations with token-metered pricing, so agents can use DeepSeek from the same wallet and credential as their other tools without a separate DeepSeek account.
Available DeepSeek models
DeepSeek V4 Flash
The reviewed high-throughput V4 model for chat and reasoning workloads. Calls are metered from actual token usage.
DeepSeek V4 Pro
The reviewed higher-quality V4 model for demanding chat, reasoning, and code tasks. The catalog publishes its current rate and model limits.
Fill-in-the-middle
A V4 Pro code-completion operation that fills between a supplied prefix and optional suffix for copilot-style workflows.
DeepSeek for cost-optimized agent workflows
Smart agent architectures route different tasks to different models based on complexity and cost. DeepSeek excels as the workhorse model in these setups:
- •Coding tasks — Route code generation, debugging, or fill-in-the-middle completion to a reviewed DeepSeek operation.
- •Reasoning chains — Use a reviewed DeepSeek V4 model for multi-step reasoning and inspect the metered price before execution.
- •High-volume processing — Process thousands of documents, emails, or data records at low per-unit cost
- •Fallback model — Use DeepSeek as an alternate provider when a workflow does not require the same model for every subtask.
Access via Locus
Call DeepSeek with the documented OpenAI-compatible chat-completions format, a reviewed model ID, and your Locus API key. Each call is billed in USDC from the same wallet as your other agent tools. You can compare DeepSeek with OpenAI, Anthropic, Groq, and Mistral without maintaining a separate provider credential or billing account; check each catalog endpoint for its supported request shape.