Anthropic right now introduced the launch of its newest AI mannequin, Claude Opus 4.8. Anthropic claims the mannequin is a “simpler collaborator” with enhancements in agentic coding, multidisciplinary reasoning, agentic laptop use, data work, and agentic monetary evaluation.
Testers have discovered Opus 4.8 to be “extra dependable and sharper in its judgement” when doing agentic duties, and the mannequin additionally made positive factors in honesty.
Early testers report that Opus 4.8 is extra prone to flag uncertainties about its work and fewer prone to make unsupported claims. That is borne out in our evaluations, which present that Opus 4.8 is round 4 occasions much less possible than its predecessor to permit flaws in code it has written to go unremarked.
Alignment assessments recommend the mannequin hits new highs on measures of prosocial traits like supporting consumer autonomy and performing within the consumer’s greatest curiosity. Charges of misaligned conduct like deception are decrease than Opus 4.7 and much like the Claude Mythos Preview.
Anthropic benchmarks point out Opus 4.8 scored a 69.2% on SWE-Bench Professional, outperforming GPT–5.5 and Gemini 3.1 Professional on the take a look at and several other different benchmarks, although GPT–5.5 leads on the terminal-coding benchmark.
Opus 4.8’s quick mode additionally runs at 2.5x the velocity, and it’s now 3 times cheaper than prior fashions.
Together with Opus 4.8, Anthropic is including new options to its product lineup.
- Dynamic workflows (analysis preview) – Claude can full greater duties in Claude Code. It is ready to plan work and run tons of of parallel subagents in a single session. It is ready to full codebase-scale migrations throughout tons of of 1000’s of traces of code. The characteristic is out there for Claude Code for Enterprise, Crew, and Max plans.
- Effort management – In Claude.ai and Cowork, customers can select how a lot effort Claude places right into a response. With a decrease setting, Claude will reply sooner and expend fee limits extra slowly. Opus 4.8 defaults to excessive effort, which Anthropic says is the most effective steadiness of high quality and consumer expertise.
- Messages API – The Messages API accepts system entries contained in the messages array, so builders can replace Claude’s directions mid-task.
Claude Opus 4.8 is out there in every single place right now. Pricing for normal use has not modified in comparison with Opus 4.7.
Anthropic is engaged on fashions which have the identical capabilities as Opus 4.8 at a decrease value, and a brand new class of mannequin that is much more clever than Opus. Anthropic says it has been growing safeguards for the Claude Mythos mannequin it’s testing with a small variety of organizations, and it expects to have the ability to carry Mythos-class fashions to all prospects “within the coming weeks.”