Anthropic ships Claude Haiku 5.5, priced to match GPT-6 Luna

Haiku 4.5 had been Anthropic's small model since October 2025, while OpenAI's GPT-6 Luna, Gemini 3.8 Flash, GLM-5.3 Flash and DeepSeek V4.1 Flash competed at the low-cost end. The Haiku 5.5 launch came with coordinated moves: the same-day Sonnet 5.5 cache-read cut, which Anthropic says makes Sonnet 5.5 about 20% cheaper on most long-running or agentic work, and monthly Claude Platform API credits for Max 5x ($100), Max 20x ($200) and Team (up to $500, pooled) subscribers.

Anthropic positions the model as a subagent paired with Opus 5.5 or Sonnet 5.5, for high-volume, cost-sensitive work such as summaries, compactions and database queries. Partner rollouts on day one included Cursor, GitHub Copilot in VS Code, Devin and Agent Arena. Devin reported 58.4% on FrontierCode 1.1, ahead of Sonnet 5 at roughly one-eighth the cost per task, and recommends it as a "sidekick" under an Opus 5.5 lead in Fusion.

Artificial Analysis's numbers come with caveats. At max effort Haiku 5.5 uses about 162k output tokens per Index task, roughly 3x GPT-6 Luna at max (~50k), so heavy token use offsets part of the per-token discount, and prompts over 100K tokens pay 5x more. That likely explains why Cursor's "10x less than Haiku 4.5" on shorter requests differs from Anthropic's "75% less" on average. Factual recall is weaker than Gemini 3.8 Flash and Luna, and a pre-release safety bug caused over-refusal on AutomationBench-AA, which Anthropic is working to fix.