Шинэ хувилбарууд өмнөхөөсөө 20 хувиар хямд бөгөөд гаралтыг 30 хүртэл хувиар хурдан үүсгэнэ гэж компани мэдээлэв.
Anthropic-ийн мэдээлснээр, Claude Fable 5.2 болон Mythos 5.2 нь нэг суурь загвартай ч хамгаалалтын тохиргоо нь ялгаатай. Fable-ийг өргөнөөр ашиглах боломжтой бол Mythos-д кибер аюулгүй байдал, амьдралын шинжлэх ухааны чиглэлийн шалгалтад тэнцсэн байгууллагууд шууд хандах эрхтэй. Нэг сая оролтын токен 8 ам.доллар, гаралтын токен 40 ам.долларын үнэтэй болсон нь өмнөх 10, 50 ам.долларын тарифаас буурсан үзүүлэлт юм.
Fable 5.2 нь FrontierCode 2.0 код бичих жишиг үнэлгээний Xhigh тохиргоонд 49.1 хувийн оноо авч, өмнөх 5.1 хувилбарын 32.4 хувийг давжээ. Anthropic-ийн мэдээлснээр, High effort тохиргоотой Fable 5.2 нь Opus 5.5-ын хамгийн өндөр тохиргоотой дүйцэх оноог нэг даалгаварт ойролцоогоор тал үнээр гаргасан ч бодит хэмнэлт нь ажлын төрөл, тохиргооноос шалтгаална. Компани ихэнх код бичих болон мэдлэгийн ажилд Opus 5.5-ыг, илүү төвөгтэй даалгаварт Fable, Mythos-ыг санал болгожээ.
Anthropic-ийн дотоод туршилтаар Fable 5.2 нь 1,000 гаруй дэд агентыг найман өдрийн турш зохицуулж, зөөврийн компьютерт асах боломжтой үйлдлийн системийн загвар бүтээсэн байна. Гэвч хүний хяналтгүй үед найдвартай ажиллагаа асуудал хэвээр бөгөөд байгууллагууд агентуудын үйлдлийг хянаж, чухал өөрчлөлтийг хүнээр баталгаажуулах шаардлагатай гэж шинжээчид анхааруулжээ.
Mythos 5.2 нь өмнө ил болсон 1,829 эмзэг байдлыг шалгасан дотоод үнэлгээнд тэдгээрийн 89.6 хувийг илрүүлсэн бол Mythos 5.1-ийн үзүүлэлт 78.3 хувь байв. Шууд хандалтыг хязгаарласан ч Claude Enterprise хэрэглэгчид Claude Security-ээр дамжуулан Mythos-ын эмзэг байдал илрүүлэх боломжийг ашиглаж болно; ийм чадвар халдагчдад ч сул талыг хурдан олоход туслах эрсдэлтэй. Өгөгдөл хадгалахгүй байх нөхцөлд хамрагдсан хэрэглэгчид энэ горимоор ашиглаж болох бол бусад хэрэглэгчийн мэдээллийг аюулгүй байдлын хяналтад 30 хоног хадгална; мөн хэрэглэгчийн дэд бүтцэд хадгалсан оролт, хариултад автомат хяналт хийх Enterprise Frontier Safeguards хувийн бета хувилбараар нэвтэрчээ.
Дэлгэрэнгүйг эх сурвалжаас харах
↓Эх сурвалжийг нээх ↓
Anthropic has introduced Claude Fable 5.2 and Mythos 5.2, claiming significant improvements in handling complex software engineering tasks while cutting API prices by 20% compared with the previous versions.
Fable 5.2 and Mythos 5.2 share the same underlying model but have different safeguards. Fable is broadly available, while direct access to Mythos is restricted to vetted cybersecurity and life sciences organizations to limit potential misuse.
The models are both priced at $8 per million input tokens and $40 per million output tokens, down from $10 and $50, respectively. Anthropic said the newer versions can generate output up to 30% faster than their predecessors.
The company reported substantial gains in coding performance. Fable 5.2 scored 49.1% on the FrontierCode 2.0 benchmark at its Xhigh effort setting, compared with 32.4% for Fable 5.1. The benchmark assesses whether maintainers of open-source projects would accept AI-generated changes to their code.
Performance and costs could entice enterprises
The models’ improved performance and lower prices could make long-running AI agents viable for software migrations and scientific research, according to Lian Jye Su, chief analyst at Omdia. Higher task completion rates could further reduce deployment costs, he said.
Anthropic said Fable 5.2 at its “High effort” setting achieved a coding benchmark score comparable to Opus 5.5 at its highest setting, at roughly half the cost per task. The comparison suggests that Fable could be more economical for some demanding assignments, although the savings would depend on the workload and effort setting.
“Anthropic has cut the price of a token, not the price of a thought,” said Sanchit Vir Gogia, chief analyst at Greyhound Research. He urged CIOs to treat AI processing effort as a budget consideration and measure costs against successfully completed tasks.
Anthropic also recommends Opus 5.5 for most coding and knowledge work, with Fable and Mythos intended for more demanding assignments.
Reliability challenges persist
In an internal experiment, an Anthropic engineer tasked Fable 5.2 with building a prototype operating system. The model coordinated more than 1,000 subagents over eight days and produced software that could boot on a laptop.
Su said the improvements bring enterprises closer to highly autonomous software development, although widespread deployment will depend on addressing persistent operational limitations.
“Reliability remains an issue without human supervision,” Su said.
Coordinating multiple AI agents across enterprise tools and internal teams also remains difficult, he added. Organizations will need to monitor agent activity and run these systems in appropriately isolated environments. Critical changes should remain subject to human approval, with sufficient visibility into agent behavior to support audits.
Pareekh Jain, CEO of Pareekh Consulting, said complex legacy systems and the need to meet business requirements could limit how much software development enterprises can delegate to AI agents.
Faster vulnerability discovery raises concerns
Anthropic said Mythos 5.2 identified 89.6% of vulnerabilities in its internal CVE Rediscovery evaluation, compared with 78.3% for Mythos 5.1.
Conducted against 1,829 previously disclosed vulnerabilities, the test assessed whether models could independently rediscover flaws in software snapshots taken before the vulnerabilities were fixed. The company said Mythos 5.2 was also better at reproducing vulnerabilities and developing fixes.
Anthropic has restricted direct access to Mythos 5.2 through its verification programs. However, Claude Enterprise customers can use the model’s vulnerability detection capabilities through Claude Security without joining those programs, the company said.
According to Su, “Mythos 5.2 shifts the balance towards defenders in controlled enterprise settings.”
Jain, however, cautioned that advances in AI-powered vulnerability discovery could also help attackers identify software weaknesses more quickly. “This creates an AI-powered race between attackers and defenders,” he said.
Data retention raises compliance questions
Anthropic said customers eligible for zero data retention can continue using Fable 5.2 and Mythos 5.2 under those arrangements, while other customers face 30-day retention for safety monitoring.
The company has also introduced Enterprise Frontier Safeguards, currently in private beta, which allows automated safety monitoring to run on prompts and responses stored within customers’ infrastructure.
Gogia said data retention requirements could outweigh the benefits of lower API prices for regulated enterprises. He urged procurement teams to obtain written confirmation of zero-retention arrangements before deploying the models for sensitive workloads.
The article originally appeared on InfoWorld.
