Anthropic Just Dropped Claude Opus 4.8 (Full Breakdown)

Summary
Brock Mesarich breaks down Anthropic's announcement of Claude Opus 4.8 for non-technical viewers, analyzing the official release announcement, pricing, and benchmark tables on an online whiteboard. He explains the new features—including configurable effort levels, dynamic workflows, honesty improvements, and the upcoming Claude Mythos preview—and demonstrates the effort settings in the Claude Cowork desktop interface.

What is shown

• [00:02] Digital whiteboard view where the presenter reviews Anthropic's announcement tweet, official blog post, benchmark table, and takeaway notes.
• [00:48] Anthropic's benchmark table comparing Claude Opus 4.8 against Claude Opus 4.7, GPT-5.5, and Gemini 3.1 Pro across various evaluations (agentic coding, terminal coding, multidisciplinary reasoning, computer use, knowledge work, and financial analysis).
• [01:13] The "Availability" and pricing section of Anthropic's blog post.
• [01:55] "A note on effort" section of the blog post explaining default high effort and effort modes.
• [02:58] Demonstration inside the Claude Cowork desktop application, switching from Opus 4.7 High to Opus 4.8, opening the model selector menu, and displaying available effort settings (Low, Medium, High (Default), Extra, Max) alongside the Adaptive thinking toggle.
• [03:33] Discussion of the "Honesty" section of the announcement, highlighting early tester reports.
• [04:47] Discussion of the "What's next?" section detailing Project Glasswing and the unreleased Claude Mythos Preview model.
• [06:01] Review of the "Also launching today" section covering dynamic workflows in Claude Code and Messages API updates.
• [06:50] The presenter's handwritten summary of the four main takeaways.
• [07:36] Quick walkthrough of Opus 4.8 selectable in Claude Cowork, regular Claude chat, and Claude Code menus.

Claims & numbers

• The presenter says Claude Opus 4.8 outperforms Opus 4.7, GPT-5.5, and Gemini 3.1 Pro on nearly all benchmarks shown, with the exception of GPT-5.5 scoring higher on agentic terminal coding (78.2% vs. 66.1%) [00:54].
• Pricing remains unchanged from Opus 4.7: $5 per million input tokens and $25 per million output tokens for regular usage, while fast mode costs $10 per million input tokens and $50 per million output tokens [01:29].
• Opus 4.8 defaults to "high effort" across tasks [02:19].
• Claude Code users can select "extra" (xhigh) or "max" effort levels, and rate limits in Claude Code have been increased to accommodate higher token usage [02:41, 02:47].
• According to Anthropic's evaluations, Opus 4.8 is roughly four times less likely than its predecessor to allow flaws in code it writes to pass unremarked [04:22].
• Project Glasswing is currently granting a small number of organizations preview access to "Claude Mythos Preview" for cybersecurity work, with wider availability expected in the coming weeks [05:25, 05:51].
• The new "Dynamic workflows" feature in research preview allows Claude Code to plan and run hundreds of parallel subagents in a single session [06:20].
• The Messages API now accepts system entries inside the messages array [06:44].
• The presenter characterizes the overall upgrade as a modest, marginal improvement rather than a game-changer [07:18].

Notable quotes

• [00:10] "I'm going to make a no-BS breakdown on exactly what's different. If you're non-technical, I'm not going to talk benchmarks and complicate this..."
• [04:48] "Users will find Opus 4.8 to be a modest but tangible improvement on its predecessor."
• [07:22] "This is definitely not a game-changing release. Of course, it is a new level up from Claude Opus 4.7, but I think this is kind of laying the groundwork for a bigger model release..."

Assessment
This is a third-party review and commentary video by an independent creator covering Anthropic's Claude Opus 4.8 launch. The presenter accurately references Anthropic's published release text and demonstrates the newly available effort controls within the genuine Claude Cowork UI, while giving a measured critique that the model represents an incremental step rather than a major leap.

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