Anthropic's Sonnet 4.6 Launch Overshadowed by Mounting Developer Concerns Over Ecosystem Policies

Anthropic recently launched Sonnet 4.6, an update described by artificial analysis as a significant leap in intelligence, reportedly placing its capabilities close to the more expensive Opus 4.6, with particular strength in ‘convex’ tasks. Historically positioned as Anthropic’s mid-tier offering, balancing cost and performance between Haiku and Opus, the new Sonnet iteration presents a nuanced cost profile. While its per-token pricing remains lower, increased token usage during inference means Sonnet 4.6 can be only marginally cheaper (around 25%) than Opus for comparable workloads.

However, the technical merits of Sonnet 4.6 are largely overshadowed by escalating developer concerns regarding Anthropic’s ecosystem management and developer relations. Critics point to the substantial subsidization of Claude Code subscriptions, which offer up to 15 times more inference usage than what API pricing would entail, potentially disadvantaging third-party businesses reliant on Anthropic’s API. A recent policy change prohibiting the use of Claude.AI OAUTH tokens for third-party products, including those built with the Agent SDK, has ignited further controversy, with developers alleging it contradicts prior guidance and restricts innovative integrations. The broader developer community also highlights Anthropic’s perceived lack of transparency in handling account bans, contested integrity of benchmark results, limited engagement with open-source initiatives, and contentious brand partner compensation practices, drawing sharp contrasts with the more collaborative and transparent approach observed from competitors like OpenAI.