DeepSeek 开发者动态：CodeWhale v0.8.49 已上线，DeepSeek 官方文档也补上了 Claude Desktop 模型映射说明

今天可发布的是一条双信号开发者更新：DeepSeek TUI 已前进到 CodeWhale v0.8.49，而 DeepSeek 官方 Anthropic API 文档现在明确写出了 Claude 风格模型名到 DeepSeek Pro / Flash 的映射关系。

中文摘要

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 a compact multi-signal developer update across two of the three tracked workflow surfaces: DeepSeek TUI and DeepSeek in Claude Code / Claude Desktop. The local deployment track was checked again, but no stronger guide-shaping delta appeared.

1. DeepSeek TUI: CodeWhale v0.8.49 is now latest

GitHub Releases now marks v0.8.49 as latest on the canonical Hmbown/CodeWhale repository, dated June 1, 2026.
The rename remains official and time-bounded: legacy deepseek and deepseek-tui binaries still ship only as deprecation shims for one release cycle and are still scheduled to disappear in v0.9.0.
README and release examples now need to be read together: the current README still treats ~/.codewhale/config.toml as the primary config home and keeps ~/.deepseek/config.toml as legacy fallback, but the v0.8.49 release page's Docker example mounts ~/.deepseek into the container. That means setup pages should explain the split instead of claiming every upstream surface already converged.
The README also now freezes a cost-tracking fact that matters to TUI users: it says the DeepSeek pricing table baked the previous promotional discount into a one-quarter base-rate adjustment as of 2026-05-31 15:59 UTC, and that the TUI cost estimator already uses the current rates.

2. DeepSeek in Claude Code / Claude Desktop: official Anthropic mapping is now documented

DeepSeek's official Claude Code guide still documents the environment-variable baseline: ANTHROPIC_BASE_URL=https://api.deepseek.com/anthropic, deepseek-v4-pro[1m], deepseek-v4-flash, and CLAUDE_CODE_SUBAGENT_MODEL=deepseek-v4-flash.
The stronger new change is in DeepSeek's official Anthropic API guide: it now explicitly documents model mapping for Claude-style names, stating that claude-opus* maps to deepseek-v4-pro, while claude-haiku* and claude-sonnet* map to deepseek-v4-flash.
DeepSeek also now says this mapping can bypass model-name restrictions in the new Claude Desktop app's developer mode when users change only the base URL and API key. That shifts this topic from a pure community workaround story to an official compatibility behavior that still needs local verification.
The migration deadline remains important: DeepSeek's quick-start still marks deepseek-chat and deepseek-reasoner for deprecation on 2026-07-24, so Claude-related setup pages should keep moving readers to exact V4 model strings.

3. 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: Docker Model Runner plus quantization browsing for llama.cpp, Ollama, and LM Studio. No stronger official or reproducible Mac-specific delta appeared today.
Release popularity alone: GitHub star count is higher again, but stars only matter when tied to a real install, runtime, or documentation change.
Loose Desktop chatter without exact docs: community notes remain useful, but today's safe Claude/Desktop delta is the official Anthropic mapping text, not weaker reposts of the same behavior.

Editorial takeaway

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