Blog / Service updates

deepseek-v4-flash Retired: Old Name Rerouted, Move to deepseek-flash

2026-10-19cdnping editors4 min read

deepseek-v4-flash was retired on September 10, 2026, when V4.1-Flash launched. The old name still works for now, but only as a compatibility layer: requests are routed to the new model and billed at Flash pricing. In your code, switch to deepseek-flash.

Everything below is current as of October 2026 and follows the official documentation. Source links are at the end.

#What exactly happened on September 10?

According to DeepSeek's September 10 announcement:

• DeepSeek-V4.1-Flash was released. To call it, set the model name to deepseek-flash.
• V4-Flash and V4-Flash-Vision-Exp were retired.
• The two old names, deepseek-v4-flash and deepseek-v4-flash-vision-exp, are temporarily routed to V4.1-Flash.
• The new pricing took effect at 04:00 UTC on September 10.

The pricing page currently says these two old names are "still usable," that the underlying models are retired, and that requests are handled by the V4.1-Flash family and billed at Flash pricing. "Temporarily" comes with no end date, so don't treat the old names as long-term.

#Is deepseek-v4-pro still available?

The official statements differ over time here, so read carefully.

The September 10 announcement said that from 04:00 UTC on September 14, deepseek-v4-pro requests would be routed to V4.1-Flash until V4.1-Pro goes live.

The changelog later states that the V4 Pro API continues to be served after September 14, with billing unchanged. As of October, the pricing page maps deepseek-v4-pro to DeepSeek-V4-Pro-0813. There is no officially announced release date for V4.1-Pro.

Bottom line: deepseek-v4-pro works today, but if you depend on it, check the original source regularly instead of relying on secondhand summaries. That includes this article.

#What is the difference between deepseek-flash and deepseek-v4-flash?

Note that deepseek-flash is a name that tracks official updates. The pricing page currently maps it to DeepSeek-V4.1-Flash-1017. The weights were updated once already on October 17, with no change to usage or pricing; see the V4.1-Flash-1017 update explained. In other words, after you switch to the new name, outputs may change between versions, so run regression tests regularly.

#How do I migrate my code?

#Step 1: Find every hardcoded model name

grep -rnE "deepseek-v4-flash|deepseek-v4-flash-vision-exp|deepseek-v4-pro" \
  --include="*.py" --include="*.js" --include="*.ts" --include="*.json" \
  --include="*.yaml" --include="*.yml" --include="*.toml" --include="*.env*" .

Don't forget CI variables, config tables in databases, and model names entered in third-party tools (editor plugins, Cherry Studio, and the like). grep won't find those.

#Step 2: Replace the names and centralize the config

Change deepseek-v4-flash to deepseek-flash. Change deepseek-v4-flash-vision-exp to deepseek-flash as well (the pricing page notes that this model supports vision). While you're at it, put the model name in an environment variable so the next rename is a one-line change. If you have a clear need for deepseek-v4-pro, keep it; don't switch to Flash by default.

#Step 3: Check which models your account can see

curl -s https://api.deepseek.com/models \
  -H "Authorization: Bearer $DEEPSEEK_API_KEY"

deepseek-flash should appear in the returned list. For the exact response format, refer to the official documentation.

#Step 4: Run regression tests

• Fix a representative set of prompts and compare the output before and after the change, checking that the format (JSON structure, tool calls) stays stable.
• Check thinking mode: the pricing page says both models support thinking and non-thinking modes, with thinking on by default. If you relied on the default behavior before, set it explicitly.
• Compare your bill. The old names are billed at Flash pricing, so costs generally shouldn't go up, but see the peak and off-peak pricing explainer for how time-of-day pricing is calculated.

#How do I avoid being caught off guard by changes like this?

Model renames, retirements, and forwarding have happened at every major provider, for example Gemini's automatic routing after deprecation and Codex retiring its old default model. A few things you can do:

1. Subscribe to cdnping's site changes tracker at /en/changes. It monitors official changelogs and pricing pages, including DeepSeek's, records additions and deletions line by line, and offers an RSS feed.
2. Use model monitoring at /en/models and the DeepSeek API status page to check availability.
3. Watch for billing changes on new features, such as pricing for the Search & Fetch API.

#Sources

• DeepSeek September 10 announcement
• DeepSeek API changelog
• DeepSeek models and pricing
• DeepSeek API docs home