Anthropic Launches Claude Opus 4.8 With Gains in Coding and Honesty

Anthropic today announced the motorboat of its latest AI model, Claude Opus 4.8. Anthropic claims the exemplary is simply a "more effectual collaborator" with improvements successful agentic coding, multidisciplinary reasoning, agentic machine use, cognition work, and agentic fiscal analysis.


Testers person recovered Opus 4.8 to beryllium "more reliable and sharper successful its judgement" erstwhile doing agentic tasks, and the exemplary besides made gains successful honesty.
Early testers study that Opus 4.8 is much apt to emblem uncertainties astir its enactment and little apt to marque unsupported claims. This is borne retired successful our evaluations, which amusement that Opus 4.8 is astir 4 times little apt than its predecessor to let flaws successful codification it has written to walk unremarked.
Alignment assessments suggest the exemplary hits caller highs connected measures of prosocial traits similar supporting idiosyncratic autonomy and acting successful the user's champion interest. Rates of misaligned behaviour similar deception are little than Opus 4.7 and akin to the Claude Mythos Preview.

Anthropic benchmarks bespeak Opus 4.8 scored a 69.2% connected SWE-Bench Pro, outperforming GPT–5.5 and Gemini 3.1 Pro connected the trial and respective different benchmarks, though GPT–5.5 leads connected the terminal-coding benchmark.

Opus 4.8's accelerated mode besides runs astatine 2.5x the speed, and it is present 3 times cheaper than anterior models.

Along with Opus 4.8, Anthropic is adding caller features to its merchandise lineup.

• Dynamic workflows (research preview) - Claude tin implicit bigger tasks successful Claude Code. It is capable to program enactment and tally hundreds of parallel subagents successful a azygous session. It is capable to implicit codebase-scale migrations crossed hundreds of thousands of lines of code. The diagnostic is disposable for Claude Code for Enterprise, Team, and Max plans.


• Effort control - In Claude.ai and Cowork, users tin take however overmuch effort Claude puts into a response. With a little setting, Claude volition respond faster and usage up complaint limits much slowly. Opus 4.8 defaults to precocious effort, which Anthropic says is the champion equilibrium of prime and idiosyncratic experience.


• Messages API - The Messages API accepts strategy entries wrong the messages array, truthful developers tin update Claude's instructions mid-task.

Claude Opus 4.8 is disposable everyplace today. Pricing for regular usage has not changed compared to Opus 4.7.

Anthropic is moving connected models that person the aforesaid capabilities arsenic Opus 4.8 astatine a little cost, and a caller people of exemplary that's adjacent much intelligent than Opus. Anthropic says it has been processing safeguards for the Claude Mythos exemplary it is investigating with a tiny fig of organizations, and it expects to beryllium capable to bring Mythos-class models to each customers "in the coming weeks."
This article, "Anthropic Launches Claude Opus 4.8 With Gains successful Coding and Honesty" archetypal appeared connected MacRumors.com

Discuss this article successful our forums