Overview
- Haiku 4.5 is available now across Claude apps and Claude Code, the API, Amazon Bedrock, and Google Cloud Vertex AI as a drop‑in replacement for Haiku 3.5 and, in many cases, Sonnet 4.
- Anthropic reports similar coding performance to Sonnet 4 at roughly one‑third the cost and more than twice the speed, positioning the model for real‑time assistants, support automation, and pair programming.
- Company benchmarks cite 73.3% on SWE‑bench Verified and roughly 41% on Terminal‑Bench, placing Haiku 4.5 near Sonnet 4 and competitive with GPT‑5 and Gemini 2.5 on those tests, with methods disclosed for replication.
- The model carries an internal ASL‑2 designation with lower measured misalignment rates than Sonnet 4.5 and Opus 4.1, reflecting Anthropic’s assessment of reduced safety risks.
- Early external tests show materially lower latency in real tasks, and Anthropic promotes a pattern where Sonnet 4.5 plans complex work while multiple Haiku 4.5 instances execute subtasks in parallel.