Uploaded May 2026 | Updated September 2026, 2 weeks ago
🚀 The LangChain 10 Days FREE Bootcamp is live: 10 lessons, free AI models only, from your first API call to a production grade RAG agent. Start with Day 0 for the roadmap and setup.
📺 Full playlist: youtube.com/watch?v=KJ3_NExk7-Q&list=PLW4pPr9JCovI&index=1
----------
Claude Opus 4.8 just dropped — and I tested it within hours of launch. Honest verdict: NOT groundbreaking. Just a slightly better Opus 4.7. And it is painfully slow — 5 to 10 minutes per prompt. Gemini 3.5 Flash would have built hundreds of apps in the same time.
Tested live in Claude AI Desktop with 2 coding prompts — 4-Cylinder Engine Simulator and Subway Surfer endless runner — with real token tracking on my Pro account.
Prompt Link:
kgptalkie.com/tutorials/generative-ai/advanced-llm-benchmark-prompt-suite
⏱️ TIMESTAMPS
0:00 Intro — Opus 4.8 launched, honest first impressions
0:20 Verdict upfront — NOT groundbreaking, slightly better than 4.7
0:32 Anthropic's own claim — built on Opus 4.7, minor upgrade
0:57 Speed problem — 5 to 10 min per prompt
1:49 Prompt 1: 4-Cylinder Engine Simulator (HTML/CSS/Vanilla JS)
2:52 Thinking took 2 minutes before writing a single line
3:34 Still no output at 4 minutes — 4.7 was faster
4:05 Engine result — RPM control, 4 states, torque panel
5:14 Token check — 10% of session used for just one prompt!
5:43 Prompt 2: Subway Surfer Endless Runner
6:15 Thinking alone: 3 minutes
6:50 Gemini 3.5 Flash comparison — 100s of apps vs 1 game
7:57 Runner demo — coins, shields, magnet, lanes
8:41 Works but nothing special
9:06 We need seconds, not 10-20 minutes
9:19 Final verdict — honest wrap-up
9:25 Outro
🔗 RESOURCES
Claude AI Desktop — claude.ai
All prompts — KGPtalkie.com
📌 WHAT YOU'LL LEARN
✅ Honest Opus 4.8 first impressions — tested within hours of launch
✅ Is it really better than 4.7? Real output shown
✅ Actual speed — how long per prompt on Claude Pro
✅ Token cost — 10% of your session per prompt
✅ Why Gemini 3.5 Flash still wins on speed
🚀 The LangChain 10 Days FREE Bootcamp is live: 10 lessons, free AI models only, from your first API call to a production grade RAG agent. Start with Day 0 for the roadmap and setup.
📺 Full playlist: youtube.com/watch?v=KJ3_NExk7-Q&list=PLW4pPr9JCovI&index=1
----------
Claude Opus 4.8 just dropped — and I tested it within hours of launch. Honest verdict: NOT groundbreaking. Just a slightly better Opus 4.7. And it is painfully slow — 5 to 10 minutes per prompt. Gemini 3.5 Flash would have built hundreds of apps in the same time.
Tested live in Claude AI Desktop with 2 coding prompts — 4-Cylinder Engine Simulator and Subway Surfer endless runner — with real token tracking on my Pro account.
Prompt Link:
kgptalkie.com/tutorials/generative-ai/advanced-llm-benchmark-prompt-suite
⏱️ TIMESTAMPS
0:00 Intro — Opus 4.8 launched, honest first impressions
0:20 Verdict upfront — NOT groundbreaking, slightly better than 4.7
0:32 Anthropic's own claim — built on Opus 4.7, minor upgrade
0:57 Speed problem — 5 to 10 min per prompt
1:49 Prompt 1: 4-Cylinder Engine Simulator (HTML/CSS/Vanilla JS)
2:52 Thinking took 2 minutes before writing a single line
3:34 Still no output at 4 minutes — 4.7 was faster
4:05 Engine result — RPM control, 4 states, torque panel
5:14 Token check — 10% of session used for just one prompt!
5:43 Prompt 2: Subway Surfer Endless Runner
6:15 Thinking alone: 3 minutes
6:50 Gemini 3.5 Flash comparison — 100s of apps vs 1 game
7:57 Runner demo — coins, shields, magnet, lanes
8:41 Works but nothing special
9:06 We need seconds, not 10-20 minutes
9:19 Final verdict — honest wrap-up
9:25 Outro
🔗 RESOURCES
Claude AI Desktop — claude.ai
All prompts — KGPtalkie.com
📌 WHAT YOU'LL LEARN
✅ Honest Opus 4.8 first impressions — tested within hours of launch
✅ Is it really better than 4.7? Real output shown
✅ Actual speed — how long per prompt on Claude Pro
✅ Token cost — 10% of your session per prompt
✅ Why Gemini 3.5 Flash still wins on speed
