What breaks if I don't migrate before the cutoff?

Any request to the DeepSeek API using deepseek-chat or deepseek-reasoner returns an error after 2026-07-24 15:59 UTC. There is no grace period and no automatic fallback to a newer model — the call fails outright, so apps, scripts, and scheduled jobs that still use the old names will stop working.

Where did deepseek-reasoner map before, and what replaces it?

deepseek-reasoner was the thinking-mode alias for DeepSeek V4-Flash. The like-for-like replacement is deepseek-v4-flash with thinking mode enabled. Moving to deepseek-v4-pro is an upgrade rather than a swap — V4-Pro is a larger, more expensive model with different reasoning behavior.

How much does deepseek-v4-flash cost after the migration?

deepseek-v4-flash is billed at $0.14 per 1M input tokens on cache miss (and $0.0028 per 1M on cache hit), plus $0.28 per 1M output tokens, with a 1M-token context window. That is the same base pricing the legacy aliases were already routed through, so a straight swap does not raise your bill.

Do I only need to change the model string?

Usually yes — update the model name in application code, gateway configs, environment variables, and evaluation suites. DeepSeek also recommends retesting thinking-mode output parsing after the switch, since prompts written against the old deepseek-reasoner may depend on its exact response format.

Is deepseek-v4-pro a drop-in upgrade from V4-Flash?

No. deepseek-v4-pro is roughly 3× the price of V4-Flash ($0.435 per 1M input, $0.87 per 1M output on cache miss), and thinking mode on V4-Pro burns dramatically more output tokens than V4-Flash-thinking did. Treat it as a separate capability tier, not a like-for-like swap.

DeepSeek · 2026-07-24 · major

DeepSeek retires deepseek-chat and deepseek-reasoner — API aliases die today

DeepSeek retires its legacy API aliases at 15:59 UTC. Apps still calling deepseek-chat or deepseek-reasoner will fail; migrate to deepseek-v4-flash or deepseek-v4-pro.

DeepSeek API social card announcing V4 model availability

DeepSeek's legacy API aliases go dark at 15:59 UTC — migrate apps to V4-Flash or V4-Pro before the cutoff.

Key specs

Cutoff	2026-07-24 15:59 UTC
Migrate to	deepseek-v4-flash / deepseek-v4-pro

Quick facts

Maker	DeepSeek
Cutoff	2026-07-24 15:59 UTC
Retired names	deepseek-chat, deepseek-reasoner
Default replacement	deepseek-v4-flash
V4-Flash price	$0.14 in / $0.28 out per 1M tokens
V4-Pro price	$0.435 in / $0.87 out per 1M tokens
Context window	1M tokens

Pricing

V4-Flash input (cache miss)	$0.14 / 1M tokens
V4-Flash input (cache hit)	$0.0028 / 1M tokens
V4-Flash output	$0.28 / 1M tokens
V4-Pro input (cache miss)	$0.435 / 1M tokens
V4-Pro input (cache hit)	$0.003625 / 1M tokens
V4-Pro output	$0.87 / 1M tokens

source ↗

What is it?

DeepSeek retires its legacy deepseek-chat and deepseek-reasoner API aliases at 2026-07-24 15:59 UTC. After the cutoff, any request that names either model returns an error — no grace period, no fallback to a newer model.

How does it work?

The two aliases were compatibility shims routing to the same underlying V4-Flash model in different modes: deepseek-chat mapped to V4-Flash non-thinking, deepseek-reasoner mapped to V4-Flash thinking. After the sunset, callers must name the target model directly — deepseek-v4-flash for the default path or deepseek-v4-pro for the larger tier.

Why does it matter?

Every app, script, gateway config, or evaluation suite that still uses the old names will 4xx as soon as the DeepSeek cutoff passes. Teams that ignored the three-month grace window now have to ship a code change to keep their DeepSeek traffic alive.

Who is it for?

developers using the DeepSeek API

Frequently asked questions

What breaks if I don't migrate before the cutoff?: Any request to the DeepSeek API using deepseek-chat or deepseek-reasoner returns an error after 2026-07-24 15:59 UTC. There is no grace period and no automatic fallback to a newer model — the call fails outright, so apps, scripts, and scheduled jobs that still use the old names will stop working.
Where did deepseek-reasoner map before, and what replaces it?: deepseek-reasoner was the thinking-mode alias for DeepSeek V4-Flash. The like-for-like replacement is deepseek-v4-flash with thinking mode enabled. Moving to deepseek-v4-pro is an upgrade rather than a swap — V4-Pro is a larger, more expensive model with different reasoning behavior.
How much does deepseek-v4-flash cost after the migration?: deepseek-v4-flash is billed at $0.14 per 1M input tokens on cache miss (and $0.0028 per 1M on cache hit), plus $0.28 per 1M output tokens, with a 1M-token context window. That is the same base pricing the legacy aliases were already routed through, so a straight swap does not raise your bill.
Do I only need to change the model string?: Usually yes — update the model name in application code, gateway configs, environment variables, and evaluation suites. DeepSeek also recommends retesting thinking-mode output parsing after the switch, since prompts written against the old deepseek-reasoner may depend on its exact response format.
Is deepseek-v4-pro a drop-in upgrade from V4-Flash?: No. deepseek-v4-pro is roughly 3× the price of V4-Flash ($0.435 per 1M input, $0.87 per 1M output on cache miss), and thinking mode on V4-Pro burns dramatically more output tokens than V4-Flash-thinking did. Treat it as a separate capability tier, not a like-for-like swap.

Try it

Change deepseek-chat → deepseek-v4-flash (non-thinking) or deepseek-v4-pro (thinking) in your API calls.