TechGlobal.news

Anthropic launches Claude 3.7 Sonnet with lower pricing and EU availability on Vertex AI and Bedrock

Enzo Reno ·

Anthropic on Tuesday released Claude 3.7 Sonnet, a new mid tier generative model for enterprise builders, with a 50 percent cut to input token pricing and expanded availability on Amazon Bedrock and Google Cloud Vertex AI in European regions. The company said the model is live today in the United States and the European Union, and is aimed at lowering inference costs while improving tool use latency for production workloads.

The release lands as enterprises lock in fourth quarter budgets and press vendors on price, latency, and compliance. OpenAI, Google, Cohere, and Meta aligned providers have all trimmed rates this year, and enterprises want multi cloud access with data residency controls.

Pricing and performance

Anthropic lists Claude 3.7 Sonnet at 1.50 dollars per million input tokens and 7.50 dollars per million output tokens via the Anthropic API, with the same rates targeted on Bedrock and Vertex AI subject to provider fees. That compares to OpenAI’s posted GPT 4o rates of 5 dollars per million input tokens and 15 dollars per million output tokens, which puts Sonnet at a lower list price for many back end tasks.

The model supports a 200,000 token context window for general use, with a one million token batch mode in preview for long document analysis according to Anthropic. On internal evaluations the company reports 88.1 on MMLU, 82.4 on GSM8K without chain of thought, 85.3 on HumanEval, and 62.0 on MMMU, with median latency below 300 milliseconds for short function calling responses at default settings.

Enterprise control and compliance

Anthropic said Claude 3.7 Sonnet ships with expanded system prompts, configurable tool use policies, and API level audit logs. Vertex AI adds EU data residency in Frankfurt and Madrid, along with VPC Service Controls and CMEK, while Amazon Bedrock integrates CloudWatch logs, Guardrails for Bedrock, and customer managed KMS keys.

The company is aligning documentation to the EU AI Act with risk class tagging, model cards, and post deployment monitoring templates. Google Cloud and AWS said the model is available in their respective marketplaces today, with standard enterprise support terms and 99.9 percent SLA for managed endpoints.

Anthropic has raised capital from Amazon and Google, including Amazon’s 4 billion dollar commitment announced in 2023 and a 2 billion dollar convertible note from Google reported in 2023. LinkedIn data indicates Anthropic’s headcount has passed one thousand employees, reflecting sustained investment in model training and safety.

For buyers, the move reinforces a price down, quality up cycle and a push to multi cloud. With AWS, Azure, and Google Cloud controlling most infrastructure spend, Synergy Research pegged their combined share above 65 percent in mid 2024, enterprises will continue to calibrate model choice by region, compliance, and total cost rather than vendor brand.