Anthropic has quietly released Cloud 3.5 Haiku artificial intelligence (AI) model to users. On Thursday, many netizens started posting about the model’s availability in the cloud’s web interface and mobile apps. Anthropic said the new generation of Haiku is the company’s fastest-growing language model. Moreover, in many benchmarks, the Foundation model even outperforms the Cloud 3 Opus, the most capable model of the previous generation. Specifically, all cloud users will have access to Cloud 3.5 Haiku, regardless of their subscription.
Anthropic releases Cloud 3.5 Haiku
While the AI firm made no announcement regarding the release of the new Haiku model, several users on X (formerly known as Twitter) posted about its availability on both the website and mobile app. Gadgets 360 staff members were also able to independently verify that Cloud 3.5 Haiku is now the default language model on chatbots. Additionally, this is the only model available to those on the free cloud tier.
Anthropic first announced the Cloud 3.5 family of AI models in October, when the first version of 3.5 Sonnet was released. At the time, the company highlighted that the 3.5 Haiku is its fastest model. Some of the upgrades in the new generations include lower latency (better response time), better instruction following, as well as precise device usage.
For enterprises, the AI firm highlighted that Cloud 3.5 Haiku excels in user-facing products, specialized sub-agent functions, and generating personalized experiences from large amounts of data.
In terms of performance, the new Haiku model scored 40.6 percent on the Software Engineering (SWE) benchmark, which outperforms the first iteration of SONET 3.5 and OpenAI’s GPT-4O. It also outperforms GPT-4o Mini on human-level and graduate-level Google-Proof Q&A (GPQA) benchmarks.
Specifically, earlier this month, Anthropic optimized Cloud 3.5 Haiku for the AWS Trenium2 AI chipset and added support for latency-optimized inference in Amazon Bedrock. The company has not yet added support for Google Cloud’s Vertex AI. The new AI model can only generate text but accepts both text and images as input.
