Claude Opus 4.8 actually blew my mind...

Summary
Alex Finn reviews and demonstrates the newly released Claude Opus 4.8 from Anthropic within Claude Code desktop. He analyzes the release notes, feature additions, pricing, and benchmark performance, then tests Opus 4.8 with his standard benchmark prompt generating a 3D first-person shooter web game.

What is shown

• [00:00] Intro slide outlining Opus 4.8 key updates: benchmark performance, unchanged pricing, cheaper fast mode, hallucination reduction, dynamic workflows, and ultracode mode.
• [03:07] Excerpt from Anthropic's blog post previewing Mythos-class models coming in the next few weeks.
• [05:35] Claude Code UI demonstration showing model selection options (Opus 4.8, Opus 4.8 1M context, Sonnet 4.6, Haiku 4.5, Opus 4.7 Legacy) and effort level configurations (Low, Medium, High, Extra, Max).
• [08:43] Google Sheets benchmark tracking sheet displaying historical scores across various coding/game-generation benchmarks.
• [09:02] Entering the benchmark prompt into Claude Code: "Build me a 3D first-person shooter using threejs in a single html file. Make this game as stylistic, fun, and visually appealing as possible. Add any mechanics, powerups, and enemies you think will make the game more fun and beautiful."
• [09:46] Demonstration of Claude Code's remote control feature synced to a mobile phone interface.
• [10:29] Gameplay and visual inspection of the generated browser game titled "Neon Assault: Survive the Grid", featuring multiple enemy waves, lighting effects, combo counters, hit markers, and collectibles.
• [11:22] Logging a score of 9.1 for Opus 4.8 on the spreadsheet benchmark.

Claims & numbers

• The presenter claims Opus 4.8 beats benchmarks, ChatGPT 5.5, and all other frontier models.
• The presenter notes the base API/subscription price remained identical to Opus 4.7, making it the first release in a while without a price increase.
• The presenter states /fast mode is now 3x cheaper than it was previously (reducing from 6x more expensive than regular mode to approximately 2x more expensive).
• Anthropic claims a 4x reduction in hallucinations compared to previous models.
• Dynamic workflows allow the model to spin up between tens to thousands of sub-agents to tackle complex multi-step coding and testing tasks in parallel.
• The presenter states Mythos-class models are slated for customer release in the coming weeks according to Anthropic's blog post.
• Opus 4.8 scored 9.1 on the presenter's 3D FPS single-prompt test, ranking it above Opus 4.7 (8.8) and previous competing models.

Notable quotes

• [00:57] "It's the same cost. This is mind-blowing... this is the first release in quite a bit of time where the price didn't go up."
• [04:12] "It will now spin up between tens to thousands of sub-agents to tackle that task."
• [10:39] "These graphics are very, very nice... this is pretty nice with from the walls to the ground... to the way the gun shoots, to the way you can see hit markers on the enemies."

Assessment
This is an independent creator review and hands-on test of Anthropic's Claude Opus 4.8 in Claude Code. The single-shot HTML/Three.js game generation is demonstrated live in real time with working gameplay, though claims regarding overarching benchmark supremacy and sub-agent scale are cited directly from Anthropic announcements rather than systematically evaluated in the clip.

Described by gemini-3.8-flash on 2026-09-29 from the video's audio and frames.