DeepSeek Legacy Model Names Retire on July 24, 2026: Migration Notes
Source-based coverage of the July 24, 2026 retirement of deepseek-chat and deepseek-reasoner, with migration steps and a post-migration summary including V4-Pro vs V4-Flash selection guidance.
- Published: 2026-07-21
- Author: LongCat Community Hub editorial team
- Updated: 2026-07-28
- Tags: DeepSeek, migration, deprecation
This site is not affiliated with, endorsed by, or sponsored by DeepSeek, Meituan, or LongCat.
What is changing
The DeepSeek API documentation, accessed on 2026-07-21, states that two legacy model identifiers will become inaccessible after July 24, 2026, 15:59 UTC:
deepseek-chatdeepseek-reasoner
Before the retirement date, these two identifiers route to deepseek-v4-flash in non-thinking and thinking mode, respectively. After the deadline, they will return errors.
Who is affected
Any application, script, SDK integration, or agent configuration that uses deepseek-chat or deepseek-reasoner as the model parameter will stop working after the deadline. This includes direct API calls, OpenAI SDK integrations, Anthropic SDK integrations, and agent tools such as Claude Code, OpenCode, or GitHub Copilot that reference these identifiers.
What developers need to do
According to the cited publisher documentation, the base URL remains unchanged. Developers only need to change the model parameter:
| Before July 24 | After July 24 |
|---|---|
deepseek-chat (routes to V4-Flash non-thinking) | deepseek-v4-flash with thinking mode disabled |
deepseek-reasoner (routes to V4-Flash thinking) | deepseek-v4-flash with thinking mode enabled, or deepseek-v4-pro for stronger reasoning |
Both new model identifiers (“deepseek-v4-pro” and “deepseek-v4-flash”) support Thinking and Non-Thinking modes, a 1M token context window, and both OpenAI ChatCompletions and Anthropic API formats, per the publisher documentation.
For developers using Claude Code or other agent tools with DeepSeek, update the model identifier in the tool's configuration or environment variables. Consult the tool's own documentation for configuration details.
Post-Migration Summary (Updated 2026-07-28)
As of July 28, the referenced model ID migration has completed. Developers who updated before July 24 should not experience disruption. For those who may still encounter errors:
- Verify the model parameter is set to
deepseek-v4-flashordeepseek-v4-pro. - Check any configuration files, environment variables, or agent tools (Claude Code, OpenCode, GitHub Copilot) that may still reference legacy identifiers.
- If using
deepseek-v4-pro, confirm that thinking mode is explicitly enabled if reasoning is needed. - Review the current DeepSeek API documentation at api-docs.deepseek.com for the latest model IDs and behavior.
Choosing between V4-Pro and V4-Flash: V4-Pro offers stronger reasoning and a larger parameter budget, suitable for complex coding, analysis, and multi-step problems. V4-Flash is optimized for speed and cost, suitable for chat, quick answers, and high-throughput workloads. Both support 1M-token context and both OpenAI and Anthropic API formats.
Source notes and limitations
This page summarises the publisher migration notice published by DeepSeek in its API documentation and preview release announcement.
- Readers should verify the current model IDs, endpoint behaviour, and pricing against the DeepSeek API documentation before deploying changes.
- This site has not independently tested the migration or the behaviour of either legacy or new model identifiers.
- DeepSeek-V4 Preview, V4-Pro, and V4-Flash are not being retired. Only the legacy identifiers deepseek-chat and deepseek-reasoner are affected.
- This page is published by an independent third-party site and is not affiliated with or endorsed by DeepSeek.
The deadline and migration steps above are taken from the cited DeepSeek publisher documentation. Confirm the latest details at api-docs.deepseek.com before making changes.
Related pages
- LongCat-2.0 and DeepSeek-V4 Preview: Documented Access and Integration Differences
A source-based comparison of documented access methods and integrations for both models.
- LongCat API Quick Start
Independent summary of the documented OpenAI-compatible request flow. Useful for developers working with compatible API formats.
- LongCat Benchmark Methodology and Evidence Policy
How this site labels benchmark evidence and what must be disclosed before publishing independent results.
- About this site
Background on how this independent third-party site sources, verifies, and discloses its content.
Sources
- DeepSeek-V4 Preview Release
Publisher documentationPublished 2026-04-24Accessed 2026-07-21
Publisher announcement for DeepSeek-V4 Preview. Contains the legacy model name retirement notice and the retirement date and time.
- DeepSeek API Documentation — First API Call
Publisher documentationAccessed 2026-07-21
Publisher documentation listing the current model IDs, their deprecation status, base URLs for OpenAI and Anthropic formats, and agent integration mentions.
- DeepSeek API Changelog
Publisher documentationAccessed 2026-07-21
Publisher changelog confirming the DeepSeek-V4 launch date (2026-04-24), model IDs, and the legacy model discontinuation timeline.
Independent third-party disclosure
This page is published by an independent third-party site. It is not affiliated with, endorsed by, sponsored by, or operated by Meituan, LongCat, or any of their affiliates. The content summarizes publicly-available primary documentation and does not represent the views of any referenced organization.
Last reviewed: 2026-07-21