DeepSeek ships V4 and sets a hard Jul-24 retirement for the deepseek-chat / deepseek-reasoner model names
On April 24, 2026 DeepSeek launched V4 - deepseek-v4-flash and deepseek-v4-pro, with 1M-token context as the new default and open weights under MIT - and announced that the legacy deepseek-chat and deepseek-reasoner model names will be fully retired and inaccessible after July 24, 2026 15:59 UTC. After that cutoff, calls using the old names error with no grace period. Migration is a one-line model-name change and pricing actually improves for most callers, so this is an upgrade with a deprecation deadline rather than a downgrade - the watch item is anything still pinned to the old names.
DeepSeek did not respond inside the 30-day window. This receipt now reads as uncontested in the public record.
DeepSeek's release note tells callers to keep their base URL and just swap the model to deepseek-v4-pro or deepseek-v4-flash.
DeepSeek's release note tells callers to keep their base URL and just swap the model to deepseek-v4-pro or deepseek-v4-flash. The old deepseek-chat / deepseek-reasoner names currently remap to v4-flash's non-thinking / thinking modes for backward compatibility, but that bridge ends at the July 24 cutoff. Pricing is not the story here: v4-flash lists at $0.14/M input and $0.28/M output - below the $0.55/M output we logged for V3 in February - while v4-pro is the premium tier at $0.435/M input and $0.87/M output. The only real cost is the forced migration before the deadline; integrations that hardcode the retired names will break.
Easiest way to run a local LLM. Pair with Open WebUI for full ChatGPT replacement.
The default ChatGPT-replacement self-host. Massive ecosystem.
Permissive (MIT) open-weight model targeting ChatGPT/Claude-class chat and agentic coding without API lock-in - the strongest open option for a self-hosted stack.
Drop-in open-weight replacement for the OpenAI/Anthropic chat APIs. Flash is the locally-runnable variant; serve it behind Open WebUI.