Bolt.DIY (GPT 4o & Gemini) vs Bolt.New | Best AI Coding Tool? - Results Are Not What You Expect!

Product(s) compared

• Bolt.DIY: Open-source / local setup version where you bring your own API key; claimed free if you have an API key.
• Bolt.New: Hosted paid product; starts at ~$20/month and offers a free account (approximately ~1M tokens/day limit).

Main workflow/UI setup (as described)

Both tools:

• Let you choose a model/provider (e.g., OpenAI GPT-4o, Anthropic, Gemini, Perplexity, etc.).
• Support building apps from prompts (example prompts used: a symbol/to-do app, landing page, and a Trello-like board app).

Key features tested & results

1) Simple “to-do app” prompt

Prompt goal: Generate a working to-do app.

Bolt.New

• Generated a working to-do app.
• UX noted: full feature set in one prompt.
• Features observed: add tasks, delete tasks, and mark completed (via UI tick/check interaction).

Bolt.DIY

• Basic to-do functionality worked (add/remove).
• Missing or harder “mark completed” behavior:
  • No obvious “completed” option initially.
  • Completion worked only after clicking/tapping the task (less straightforward UX).
• UI look criticized as too basic / poor.

Unique point: Functionality was partially there in Bolt.DIY, but UX/feature clarity lagged behind Bolt.New.

---

2) Landing page prompt

Prompt: “Create a stunning looking landing page…” (with styling expectations like cards, hero section, animations, gradients)

Bolt.New

• Produced a visually richer page:
  • hero section, navbar, colors, icons
  • cards
  • animations/zoom effects, gradient effects
• Critique: uses “dummy images,” but overall presentation was good.

Bolt.DIY

• Output described as plain/basic and “terrible.”
• Follow-up attempts:
  • Tried prompts like Tailwind / React / styling enhancements.
  • Result: still broken/low quality; issues included Tailwind config not done correctly (Tailwind setup/doc config error), code rewrites, and continued poor styling.
• User concluded they were “totally done” with Bolt.DIY after repeated failures.

Unique point: Bolt.DIY repeatedly failed to implement styling correctly (especially with Tailwind), while Bolt.New delivered polished UI immediately.

---

3) Trello-like board (Kanban) app with drag-and-drop

Prompt goal: Trello-like app with multiple boards, add tasks, and drag/drop.

Bolt.New

• Functionality success:
  • boards exist
  • add board
  • add tasks
  • drag-and-drop worked (confirmed by user moving cards)
• UI:
  • Styling “off” / not great, but still functional.

Bolt.DIY

• User frustration/limitations:
  • Could not find working features for adding new tasks/boards and moving between boards.
  • Drag-and-drop was either missing or non-functional initially.
• Even after prompting to “enhance” and fix drag/drop:
  • result described as not running correctly at one point (“no preview available” / issues running npm dev).
• Even when task creation worked later, drag/drop still didn’t work.
• User concluded it was not intended to work correctly.

Unique point: Bolt.New handled the “Trello” spec (including drag/drop) much better; Bolt.DIY failed key interactive requirements.

---

Overall pros / cons mentioned

Bolt.New — Pros

• Better UI quality from prompts (landing pages looked much more polished).
• More complete functionality in tests (to-do completion, Trello-like drag/drop and multi-board operations).
• “Miles ahead” in the reviewer’s experience.
• Mentions of more advanced integrations, including option to connect to Supabase and add backend functionality.

Bolt.New — Cons

• Styling may still have issues in the Trello test (“styling looks a bit off”).
• Some bugs existed, but core functionality worked.

Bolt.DIY — Pros

• Can be free if you already have an API key.
• Runs locally and supports multiple model providers (via your chosen setup).

Bolt.DIY — Cons (dominant theme)

• Terrible UI outputs (especially compared to Bolt.New).
• Unreliable functionality:
  • missing/awkward to-do completion UX
  • Trello board features incomplete (or not working)
  • drag-and-drop not working as expected
• Technical/styling failures (Tailwind config not set correctly; repeated poor attempts).
• Sometimes preview/build issues (“no preview available”, project not running as intended).

Pricing / value comparison (explicitly stated)

• Bolt.New: starts around $20/month, with a free account (~1M tokens/day).
• Bolt.DIY: free if you bring an API key.

Comparisons / verdict from the reviewer

• The reviewer states plainly: Bolt.DIY is not a replacement for Bolt.New.
• Conclusion: Bolt.New is “miles ahead” and the reviewer recommends Bolt.New for now.
• Final recommendation line: “you’re better off with Bolt.New.”

Unique points mentioned about contributors/speakers

• No distinct multi-speaker viewpoints were clearly separated; the same reviewer voice provides all observations.

Concise verdict / recommendation

Based on tests (to-do app, landing page, Trello-like kanban with drag/drop), Bolt.New is recommended. Bolt.DIY produces noticeably worse UI and frequently fails key interactive features, so it should not be considered a replacement for Bolt.New in the scenarios tested.