88Z AI, 29 May — Anthropic has released its flagship model, Claude Opus 4.8, featuring significant enhancements in terminal engineering and knowledge work, with some metrics surpassing the unreleased Mythos. The core breakthrough lies in improved honesty, reducing code defect false negatives to one-quarter of the previous generation and overconfidence probability to one-tenth, earning high praise from the CEOs of Cursor and Devin.
At a glance
- Claude Opus 4.8 significantly improves terminal engineering and knowledge work capabilities, with some metrics surpassing the unreleased Mythos.
- Model honesty is notably enhanced, with code defect false negatives reduced to 1/4 of the previous generation and overconfidence probability down to 1/10.
The update
Anthropic has officially launched its flagship large language model, Claude Opus 4.8. The model achieves a significant leap in processing capabilities for terminal engineering and various knowledge work tasks, with some performance metrics even surpassing those of the unreleased Mythos model. The core highlight of this upgrade is the substantial improvement in model honesty. Data shows that the new model's false negative rate for code defects has been reduced to one-quarter of the previous version, while the probability of exhibiting overconfident behavior has dropped significantly to one-tenth of the original level. This progress has attracted widespread attention in the industry, with the CEOs of both Cursor and Devin giving high praise to the model's performance.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 量子位Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.