Full story
Claude Opus 4.8 Launch
Anthropic released Claude Opus 4.8 on Thursday, positioning it as an upgrade to Opus 4.7 with new capabilities and the same standard pricing as the previous Opus release.
“Anthropic has released its latest AI model with Claude Opus 4”
Alongside the model, Anthropic shipped Dynamic Workflows in Claude Code, described as a system that can plan a task and then run hundreds of parallel subagents in a single session before verifying outputs.
The launch also introduced fast mode for Opus 4.8, which runs at 2.5× speed and is priced at $10 per million input tokens and $50 per million output tokens, while standard Opus pricing remains $5 per million input tokens and $25 per million output tokens.
In Anthropic’s own launch framing, early testers reported that Opus 4.8 is “more likely to flag uncertainties about its work and less likely to make unsupported claims,” and the company said it is “around 4x less likely than its predecessor to allow flaws in code it’s written to pass unremarked.”
Dynamic Workflows and Effort
Dynamic workflows are implemented in Claude Code as a JavaScript script that orchestrates subagents at scale, with Claude writing the script and a runtime executing it in the background while the user’s session stays responsive.
The runtime applies hard limits, allowing up to 16 concurrent agents and capping each run at 1,000 agents total, while the workflow script itself cannot touch the filesystem or shell.
Anthropic also added an effort control on claude.ai and Cowork, letting users choose how much effort Claude expends, with higher settings described as making Claude “think more frequently and more deeply” and lower settings producing faster responses.
In the same launch package, fast mode is described as a high-speed configuration of Claude Opus, toggled with /fast in Claude Code, and marked by a small ↯ icon for an active session.
Honesty, Benchmarks, and Mythos
Anthropic’s release messaging centers on “honesty,” with the company saying Opus 4.8 is more likely to flag uncertainties and less likely to make unsupported claims, and that its evaluations show it is “around 4x less likely” to let code flaws pass unremarked.
“Anthropic rolls out Claude Opus 4”
On performance, The Decoder reported that Opus 4.8 scored 69.2 percent on agentic coding (SWE-Bench Pro), up from 64.3 percent for Opus 4.7 and 58.6 percent for GPT-5.5, and it also cited 49.8 percent without tools and 57.9 percent with tools on Humanity’s Last Exam.
The Decoder also said Opus 4.8 is designed to communicate uncertainties better, and it described Anthropic’s expectations for Mythos-class models as “in the coming weeks,” once safety measures are in place.
TechCrunch added that Anthropic is still holding back its most advanced Mythos model after a tentative preview raised cybersecurity concerns, but it quoted Anthropic saying it expects to bring Mythos-class models to all customers “in the coming weeks.”
