LIVE MODELV4 FLASH / AGENT ROUTER

One V4 Flash model for three coding agents

DeepSeek V4 Flash works with Claude Code through the Anthropic Messages gateway, Codex CLI through the Responses gateway, and OpenCode through OpenAI-compatible Chat Completions; each tool needs its own base URL and wire protocol.

Open the integration guide Inspect pricing

Claude Code

Messages

/anthropic

Codex CLI

Responses

/v1

OpenCode

Chat

/api

Model

V4 Flash

tool-ready

Operational brief

From signal to shipping

01 / CLAUDE CODE

Use the Anthropic-compatible route

Set ANTHROPIC_BASE_URL to the MyDeepSeek /anthropic base and authenticate with ANTHROPIC_AUTH_TOKEN. Map the Claude model environment variables to deepseek-v4-flash so subagents and main turns stay on the callable model.

Base URL: https://mydeepseekapi.com/anthropic
Wire endpoint: /anthropic/v1/messages
Native streaming and tool-use events are preserved

02 / CODEX

Configure a Responses provider

Current Codex CLI custom providers use wire_api = responses. Point the provider base_url at https://mydeepseekapi.com/v1 and keep the key in an environment variable referenced by env_key.

Provider base: https://mydeepseekapi.com/v1
Wire API: responses
Supports streamed text, tool calls, and terminal usage

03 / OPENCODE

Use an OpenAI-compatible provider

OpenCode can use its OpenAI-compatible provider with baseURL https://mydeepseekapi.com/api and model deepseek-v4-flash. The client then calls the deployed /chat/completions route.

Provider package: @ai-sdk/openai-compatible
Base URL: https://mydeepseekapi.com/api
Create a dedicated model entry for deepseek-v4-flash

Questions / answers

The short answers that unblock a build.

Can Claude Code use DeepSeek V4 Flash?+

Yes. Configure Claude Code with the MyDeepSeek Anthropic base URL, an API token, and V4 Flash model mappings.

Does Codex use the same endpoint as OpenCode?+

No. Current Codex CLI uses the Responses protocol at the /v1 base, while OpenCode uses OpenAI-compatible Chat Completions at the /api base.

Why should each coding tool have a separate key?+

Separate keys make rotation, revocation, usage attribution, and cost troubleshooting safer and clearer.

Continue the route

Related V4 Flash guides

V4 FLASH / FIELD GUIDEDeepSeek V4 Flash, decoded for production V4 FLASH / API MAPThe shortest path to a V4 Flash response V4 FLASH / COST SIGNALPrice V4 Flash with cache hits separated

Official DeepSeek sources

DeepSeek V4 Flash release notes DeepSeek API models and pricing

Ready / next action

Copy the complete verified configuration for your tool.

Open the integration guide