@deepseek_ai You can already use V4-Flash for free in Codezal β open source, bring your own API key, run it as a desktop agent. π
https://t.co/DsaekljYr0
π DeepSeek-V4-Flash is live in open beta β and the agent leap is real.
It now beats V4-Pro-Preview and speaks Responses API natively.
Best part: bring your own key. Codezal is open source β plug V4-Flash in and run it as a desktop agent. No subscription. Your data, your box.
π New in Codezal:
β¨ AI pre-commit review with one-click auto-fix β catch issues before they land
π WYSIWYG Markdown editor with Source / Visual / Preview modes
ποΈ Multi-root workspaces + token & cost budgets to keep autonomous goals in check
Ship faster, break less.
GLM 5.2 dropped today π₯
If you're on a GLM Coding Plan, it already works in Codezal β just paste your key. Same subscription, now in a native GUI with a built-in editor + LSP and Orchestra mode.
https://t.co/DsaekljYr0
New in Codezal: a full built-in code editor.
β Monaco engine (yes, the VS Code one)
β LSP: autocomplete, go-to-definition, diagnostics
β inline diff view
β git panel + file tree
Your agent and your editor, finally in one app.
https://t.co/DsaekljYr0
The Codezal plugin marketplace is just a GitHub repo.
Install curated plugins from the official marketplace β or point Codezal at your own repo.
Open. Ownable. No gatekeeper.
https://t.co/tZ87A6CYLj
Your AI client shouldn't pick your AI for you.
Codezal works with DeepSeek, Claude, OpenAI, Gemini, Mistral, Groq, Kimi β 4,968+ models total.
Bring your own API key. Nothing proxied through us.
https://t.co/DsaekljYr0
What if instead of one AI, you had 5 working in parallel?
Codezal's Orchestra mode does exactly that: one conductor LLM manages the task, up to 5 worker LLMs execute in parallel β via API or real CLI (claude, codex, opencode).
https://t.co/DsaekljYr0