Use DeepSeek V4 Flash in Claude Code: setup and pricing

Set ANTHROPIC_BASE_URL and the model to run DeepSeek V4 Flash (deepseek-v4-flash) in Claude Code. $0.09 input / $0.18 output per 1M tokens, with a config you can paste.

Create an API keyFull Claude Code guide
Model ID
deepseek-v4-flash
Input / 1M
$0.09
Output / 1M
$0.18
Context
1M

Set up Claude Code with DeepSeek V4 Flash

  1. Create an API key in the BigModel.io console and give it a spend limit; agents send long contexts.
  2. Put the block below in ~/.claude/settings.json with your key. All three model slots point at models you can call; background tasks use the haiku slot.
  3. Run claude in your project. /status shows the base URL and deepseek-v4-flash as the model.

~/.claude/settings.json

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://bigmodel.io",
    "ANTHROPIC_AUTH_TOKEN": "sk-your-bigmodel-key",
    "ANTHROPIC_MODEL": "deepseek-v4-flash",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "deepseek-v4-flash",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "deepseek-v4-flash",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "deepseek-v4-flash"
  }
}

Use ANTHROPIC_AUTH_TOKEN, not ANTHROPIC_API_KEY, and never commit this file with the key in it.

Check the key and model first

This request should return a short reply. If it fails, the tool will fail the same way.

curl https://bigmodel.io/v1/messages \
  -H "Authorization: Bearer $BIGMODEL_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{"model":"deepseek-v4-flash","max_tokens":64,"messages":[{"role":"user","content":"ping"}]}'

401 means the key is wrong, 404 the model ID; fix 401 · fix 404 · /v1/messages errors

What it costs

DeepSeek V4 Flash costs $0.09 per 1M input tokens and $0.18 per 1M output tokens, $0.03 per 1M cached input tokens. A typical agent turn with 20K tokens in and 2K out costs about $0.0022; 100 turns about $0.216. DeepSeek V4 Flash pricing →

Other setups