Official2026-05-28

DeepSeek TUI 的 CodeWhale 迁移仍在推进：v0.8.43 已成最新 release，`codewhale` 仍是规范命令

今天可发布的开发者信号仍然来自 TUI 主线：GitHub Releases 已更新到 CodeWhale v0.8.43，而本地部署与 Claude Code 两条线今天没有足够新的官方变化，不应硬写新结论。

中文摘要

今天可发布的开发者信号仍然来自 TUI 主线：GitHub Releases 已更新到 CodeWhale v0.8.43，而本地部署与 Claude Code 两条线今天没有足够新的官方变化，不应硬写新结论。

阅读提示

这篇中文稿保留原始来源链接，并把 DeepSeek 官方发布、报道和市场传闻分开标注。购买相关判断仍以 /zh/pricing 的真实库存卡片为准；出现在新闻或基准中的模型不代表可购买。

英文原文

Accepted developer update

Today's safe publishable change is on the DeepSeek TUI track. The strongest source-backed delta is small but real: the canonical GitHub release line has moved from the May 24 milestone to a newer stable tag, so pages that claim to track the current install surface should not stay pinned to v0.8.42.

What changed upstream

  • GitHub Releases now marks v0.8.43 as latest on the canonical Hmbown/CodeWhale repository.
  • The rebrand contract is still explicit: codewhale / codewhale-tui remain the canonical dispatcher/runtime pair, while legacy deepseek / deepseek-tui binaries are still temporary deprecation shims scheduled to disappear in v0.9.0.
  • The canonical Docker path advanced too: the release page now shows ghcr.io/hmbown/codewhale:v0.8.43, so guide pages should move that pinned example instead of leaving the older 0.8.42 tag in place.
  • The current README still keeps the practical developer hooks visible: skills discovery order, GitHub-installable skill packs, and self-hosted provider knobs such as SGLANG_BASE_URL and VLLM_BASE_URL remain part of the documented surface. That matters because DeepSeek TUI is still easiest to rank for when the page explains concrete developer workflows rather than only the rename story.

Why this matters for crawlable setup pages

  1. The stable reference point should now be v0.8.43, not v0.8.42.
  2. Canonical install, auth, and doctor examples should lead with codewhale, not a legacy shim command.
  3. Docker users should see the pinned ghcr.io/hmbown/codewhale:v0.8.43 example.
  4. TUI pages that mention skills or self-hosted runners can safely keep that framing, because the current README still documents both.

What we rejected today

  • DeepSeek local development: the official DeepSeek V4 Flash model card still shows the same maintained local baseline already covered on this site: vLLM and SGLang launch commands, Docker Model Runner, OpenAI-compatible curl checks, and the quantization browser for llama.cpp, Ollama, and LM Studio. No stronger official Mac- or guide-shaping delta appeared today.
  • DeepSeek in Claude Code / Cloud Code: the official DeepSeek Claude Code guide still uses the same Anthropic-compatible environment recipe already reflected on this site, including ANTHROPIC_BASE_URL=https://api.deepseek.com/anthropic, deepseek-v4-pro[1m], deepseek-v4-flash, CLAUDE_CODE_SUBAGENT_MODEL, and CLAUDE_CODE_EFFORT_LEVEL=max.
  • Community-only local runtime and Desktop chatter: useful discovery leads, but not strong enough to publish without a clearer official or reproducible milestone tied to today's three tracked surfaces.

Editorial takeaway

Today's public update belongs on the DeepSeek TUI guide, the dedicated /deepseek-tui landing page, and the news stream. It does not affect stocked plans, pricing cards, stock, or any purchasable inventory surface.

Sources checked

查看原始来源