Video summary

ChatGPT vs Claude vs Gemini vs 2 AI lainnya - Bagusan Mana??

Main summary

Key takeaways

Product Review

Product(s) discussed

This video is a comparison/review of multiple AI assistants rather than a single product. It focuses on:

  • ChatGPT
  • Claude (Cloud)
  • Gemini
  • Grok
  • DeepSeek

It also mentions several other China AIs (e.g., Kimi / Minimax, and others).

Key points, features, pros/cons, and user experience (by AI)

1) ChatGPT (GPT Chat / OpenAI)

Main features mentioned

  • A “thinking” mode for deeper reasoning.
  • Multimodal output: text + images + video + other capabilities (“things like resets”).
  • A mobile/vision-like capability: can analyze what you’re filming/doing from video (e.g., photographing a car button to identify it).
  • Voice / friend-like chatting aimed at sounding more natural.
  • Strong Memory system for remembering user preferences.
  • Toolkit-style add-ons:
    • Codex for coding
    • Separate modes for image-focused tasks, etc.
  • Mentions an option to import memory later to avoid being locked into one platform.

Pros

  • Best-in-class memory (author claims it was the first with memory).
  • Conversational feel; “thinking” for deeper answers.
  • Broad “do everything in one place” coverage.

Cons / limitations

  • The author stopped paying for the most expensive tier, saying the cost-benefit isn’t always worth it.
  • No major explicit downside beyond cost/value tradeoffs.

Verdict (from the video)

  • Best for users who want one AI that does many things—especially if you value memory and reasoning.

2) Claude / “Cloud” (Anthropic)

Main features mentioned

  • Claude Code: chat-like assistance “at the terminal,” geared toward coding/automation.
  • Focus on text + coding.
  • Ability to create artifacts (e.g., games, running pages).
  • Claude Cowork: automation embedded into apps (author says it was mainly on Windows, later also Mac).
  • More “humanistic” writing outcomes (copywriting/script style).

Pros

  • Better workflow for coding/automation (less copy-paste than some alternatives).
  • Strong for content scripts (e.g., YouTube scripts) with more human-like phrasing.
  • Quick feature releases and adaptation (as described by the author).

Cons / limitations

  • Missing image and video capabilities (unlike ChatGPT/Gemini).
  • Token usage can feel inefficient, making higher tiers less practical for automation.
  • The author suggests it initially missed major markets (not developers directly at first), though it still serves non-technical builders.

Verdict (from the video)

  • Best for coding + automation and for human-sounding writing (scripts/content).

3) Gemini (Google AI)

Main features mentioned

  • Bundled offering integrated into the Google ecosystem (example: 2 TB storage + Gemini Pro).
  • Integration with Google Drive / Sheets / Docs / Gmail and other Google products.
  • Image generation and drawing; author claims Gemini’s drawing is better than ChatGPT’s.
  • Deep Research: research mode pulling from many sources (example use with “thinking pro” / complex questions).
    • Research time mentioned: ~5–10 minutes
  • LM Notebook tool (after purchase):
    • Learning/answering grounded in your uploaded materials (PDFs, YouTube, notes, etc.)
    • Author emphasizes improved accuracy when sources are provided by the user.

Pros

  • Best “all-in-one” experience for users living in Google Drive/Docs/Sheets.
  • Strong Deep Research + source-grounded learning via LM Notebook.
  • Solid image/drawing capabilities.
  • Bundling value (author recommends “one and get all”).

Cons / limitations

  • No major downside explicitly framed; main tradeoff is simply different from ChatGPT/Claude rather than “worse.”

Verdict (from the video)

  • Best for users wanting Google ecosystem integration, research, and learning from uploaded sources, plus image/drawing strength.

4) Grok (xAI / Elon Musk)

Main features mentioned

  • Strong advantage in situational accuracy because Grok is tied to X (Twitter) data.
  • “Situational awareness”: better at the latest news, including distinguishing hot stories and hoaxes.
  • Example narrative evaluation based on evolving data (author contrasts apparent “failure” with eventual data support).
  • Author prefers Grok’s voice mode:
    • Says Grok voice articulation/type is better than ChatGPT/Gemini for voice delivery.
    • Claims other AIs aren’t “famous for voice mode” and are worse there.

Pros

  • Good for real-time/news verification and faster debunking.
  • Better voice mode (per author).

Cons / limitations

  • The author does not currently pay for Grok (no found use case).
  • Mentions that paid value is unclear for their personal needs.

Verdict (from the video)

  • Best for users who want current, X-based situational awareness and care about voice mode.

5) DeepSeek (“Deepsik” / China AI)

Main features mentioned

  • Initially popular due to being cheap with good results; later attention shifted due to competition.
  • Describes the broader “China AI” landscape:
    • DeepSeek plus others like Kimi and models from TongYi/Minimax-related groups, etc.
  • Notable feature: “Deep Think” (deeper thinking mode), tied to openness (author claims DeepSeek published the paper as open source).
  • Pricing example: about ~$3/month (author’s trial estimate).
  • Claims DeepSeek hasn’t released new models in the last few months, so it may be slightly outdated.
  • Use-case recommendation:
    • If you want to learn about China, it can sometimes be better because it’s trained with more Chinese-language textbooks/data (e.g., content about China/huawei, etc.).

Pros

  • Strong budget-friendly access.
  • Deep Think for reasoning.
  • Useful for China-focused learning and Chinese-language material.

Cons / limitations

  • Potential lag in freshness (no new models for months per author).
  • May not match the newest model quality across all categories.

Verdict (from the video)

  • Best as a cheap reasoning option, especially for China-focused learning.

Comparisons / decision framework used in the video

  • The author argues the “best AI” depends on your needs because each platform has different strengths and sometimes unique features.
  • Comparison method:
    • Give the same task to multiple assistants and compare which produces the best result.
  • Preference/style comparisons:
    • Grok: more straightforward/direct
    • ChatGPT: more “playing around,” more “framework”
    • Claude: more like “paper,” can take longer to produce (per the author’s framing)

Ratings / numerical scores mentioned

  • No explicit star ratings or formal numerical scores are provided.
  • Numerical/contextual mentions include:
    • Gemini Deep Research time: ~5–10 minutes
    • Pricing/tier references (some exact numbers are garbled in subtitles):
      • ChatGPT: author mentions tier changes (includes references like “Go”)
      • Claude: mentions $20 vs $100 tiers; later also references very expensive newer options (subtitle garbling)
      • Claude model names referenced as “Sonnet and Opus,” including “Opus 46”
      • DeepSeek: about ~$3/month in a China AI example
    • A “leaderboard” is referenced:
      • Arena with “top 1 / top 4” style guidance, but no clear verified numeric scores are shown in subtitles.

Overall pros/cons by AI (condensed)

  • ChatGPT: strongest all-rounder + memory + multimodal; top tiers may be costly.
  • Claude: strongest for coding/automation and human-sounding writing; no image/video; possible token inefficiency.
  • Gemini: best Google integration + research + source-grounded learning (LM Notebook); strong drawing/images.
  • Grok: best for latest X-based news/situational awareness; best voice mode (per author), though the author doesn’t pay.
  • DeepSeek: best budget reasoning and China-focused learning; may lag behind newest models.

Unique points / recommendations mentioned (distinct themes)

  1. AI choice is like choosing streaming services: pick the best match for your needs.
  2. A multi-AI setup can make sense because different AIs have exclusive features.
  3. ChatGPT:
    • “Thinking” mode
    • Mobile/video understanding
    • Voice/chat feel
    • Strong memory
    • Codex coding focus
    • Memory portability idea
  4. Claude:
    • Terminal automation via Claude Code
    • Artifact creation
    • Claude Cowork
    • Text/coding-only positioning
    • More humanistic writing
    • Token-cost concerns
  5. Gemini:
    • Google Drive bundle example (e.g., 2 TB)
    • Google ecosystem integration
    • Deep Research (5–10 min)
    • LM Notebook grounding in uploaded docs
    • Strong drawing/images
  6. Grok:
    • X/Twitter-based up-to-date information
    • Faster hoax detection
    • Situational awareness
    • Voice mode better than others (per author)
  7. DeepSeek / China AI:
    • Cheaper entry
    • Deep Think as a distinctive reasoning approach
    • Openness origin (paper)
    • Model freshness may lag
    • Better for Chinese-language/history content
    • Compare using “arena/leaderboards”
  8. Don’t buy annual AI plans; AI changes quickly—favor discounted shorter-term deals (around 20–30% discount mentioned).
  9. Choose based on use case (coding, content scripts, automation, China learning, etc.), then test with the same prompt across AIs.
  10. Budget testing method:
    • Use an “Open Router”-like approach to try models via top-ups (subtitle garbling).
  11. Personal conclusion:
    • The author uses different AIs depending on current work and disables unused ones to reduce cost.

Concise verdict / recommendation

Recommendation: There isn’t one “best” AI overall. Based on the video:

  • Choose ChatGPT for an all-around assistant with memory and multimodal output.
  • Choose Claude for coding/terminal automation and script/copywriting that feels more human.
  • Choose Gemini for Google ecosystem integration and source-grounded learning/research (Deep Research + LM Notebook).
  • Choose Grok for latest news/situational awareness from X and for better voice mode.
  • Choose DeepSeek for budget-friendly reasoning, especially for China-focused content.

Overall strategy: Match the AI to your primary task, and test by giving the same prompt to each model to confirm fit.

Speakers / perspectives

  • A single main speaker (William) dominates the review, sharing personal experience, pricing/tier changes, and preferred use cases per AI.
  • No other speakers are clearly distinguishable from subtitles.

Original video