Yes. DeepSeek documents deepseek-flash and deepseek-v4-pro as API model IDs, and its Responses API supports function tools for agent backends. Per the current model documentation checked on September 16, 2026, the Quick Start, model table, and Change Log state that V4 Pro API service continues after September 14. An earlier September 10 announcement had described a plan to route deepseek-v4-pro toward V4.1 Flash before V4.1 Pro; treat that as a historical note that the current pages supersede. Documentation alone does not prove the backend that handled any specific request.
The requested model ID is not proof of the model that handled a request, and neither is a complete agent. Verify the provider’s current documented mapping. Your caller or agent harness still owns repository access, command execution, permissions, state, validation, tests, and safe handling of generated function arguments. Test the complete loop on a small repository task that reads several files, makes a bounded change, runs tests, and repairs one failure.
Separate model quality from agent quality. A failed run may come from context assembly, tool permissions, prompts, execution infrastructure, or the model. Record traces and compare the same task with one variable changed at a time. Review all generated code before merge.
Use a non-sensitive repository first and set execution limits. This answer does not verify price, license, quota, local deployments, universal framework compatibility, or the backend of an actual request. Record the requested model ID, documented mapping, framework, settings, and date so the result can be reproduced.
Source: DeepSeek API Quick Start; DeepSeek Models & Pricing; DeepSeek Change Log; DeepSeek V4.1 Flash announcement; DeepSeek Responses API. Last checked: September 16, 2026 (documentation review; no API or model test).
