Before its IPO, Anthropic launched a small-scale model dubbed the "king of cost-effectiveness," with Claude Haiku 5.5 reducing operational costs by 75%.

Anthropic states that Haiku 5.5 is the company's most affordable, fastest, and most powerful small model to date, designed for high-throughput, cost-sensitive applications. Its capabilities in tasks such as programming and general computing have improved significantly compared to the previous version, 4.5. For the first time, it incorporates an adjustable inference "effort" setting. For requests with prompt tokens not exceeding 100,000, the API price is 90% lower than the prior generation; beyond that threshold, the price is reduced by 50%. Additionally, Haiku 5.5 can be used in tandem with the first two Claude 5.5 models released within the past month—Opus 5.5 and Sonnet 5.5—functioning as a "sub‑agent" to support them. The cache‑read rate for Sonnet 5.5 has been cut in half to $0.10 per million tokens.

wallstreetcnOct 8 04:36

Express News | Anthropic has released Claude Haiku 5.5, with costs reduced by approximately 75% on average.

Anthropic has released Claude Haiku 5.5, calling it the company's fastest and most powerful small model to date, primarily designed for high-frequency, cost-sensitive use cases such as summarization, classification, database querying, customer service, browser interactions, and subtasks of AI agents. Haiku 5.5 reduces average runtime costs by approximately 75% compared to Haiku 4.5, with input and output pricing of $0.10 and $0.50 per million tokens, respectively, for requests under 100,000 tokens. It also introduces adjustable inference intensity to the Haiku series for the first time. Meanwhile, Anthropic has halved Sonnet 5.5's cache read price to $0.10 per million tokens, claiming this will cut the cost of most agent-based tasks by about 20%. Haiku 5.5 is now available on the Claude Platform, AWS, Google Cloud, and Microsoft Azure.

LatestOct 8 02:13

Security Risks of AI Agents Spark Concerns; JPMorgan CEO Dimon: Anthropic’s Mythos Could Amplify Cyber Threats by Approximately 10 Times

JPMorgan CEO Jamie Dimon said in an interview that Anthropic's Mythos AI model could increase global cybersecurity risks by about tenfold.

Zhitong FinanceOct 6 22:56

South Korea's inflation rate eased to 2.9% in September, with the central bank projecting it to remain near 3% in October.

South Korea announced on October 2 that the consumer price index (CPI) rose 2.9% year-on-year in September, down 0.2 percentage points from August's 3.1%, while increasing 0.3% month-on-month. The

Sina FinanceOct 2 18:55

Anthropic takes another step toward going public: reports reveal it is aiming for an IPO valued at $2 trillion ahead of Thanksgiving, with an investor day scheduled for October 14.

According to reports, Anthropic could begin its roadshow as early as November, aiming to complete its IPO before the U.S. Thanksgiving holiday on November 26. The company will hold a pre-IPO investor day on October 14 and has already extended invitations to a group of institutional investors; some potential investors believe the firm's fair valuation could range from $1.8 trillion to $2 trillion.

WallStreetCNOct 2 03:01

Express News | Report: Anthropic Aims to Complete a Large-Scale IPO Before Thanksgiving

According to people familiar with the matter, AI model developer Anthropic PBC is seeking an IPO as early as mid-November. The company had previously postponed its listing plans. The maker of Claude could kick off its formal IPO roadshow as early as the week of November 9, potentially enabling trading to begin before Thanksgiving. With Thanksgiving falling on November 26, trading activity typically slows in the days surrounding the holiday. Sources say Anthropic still expects to complete its IPO no later than year-end. Internal discussions are ongoing, and the timeline could still be adjusted. (Bloomberg)

LatestOct 2 01:32