Anthropic has launched Claude Haiku 4.5, making what was recently frontier-level AI performance available at dramatically reduced costs and faster speeds. The new small model delivers performance comparable to Claude Sonnet 4—which was state-of-the-art just five months ago—but at one-third the cost and more than twice the speed.
Performance Meets Efficiency
Claude Haiku 4.5 represents a significant shift in the AI performance-cost trade-off. The model achieves similar coding performance to its larger predecessor while introducing substantial improvements in specific areas like computer use, where it actually surpasses Claude Sonnet 4.
This combination of high intelligence and remarkable speed makes Haiku 4.5 ideal for real-time, low-latency applications including:
- Chat assistants
- Customer service agents
- Pair programming
- Rapid prototyping
- Multi-agent orchestration
Benchmark Results
Claude Haiku 4.5 demonstrates impressive capabilities across standard AI benchmarks, positioning it as one of Anthropic's most powerful models to date despite its "small" classification. Early testing from industry partners shows remarkable results:
- Agentic Coding: Achieves 90% of Sonnet 4.5's performance in Augment's evaluation
- Response Speed: Runs 4-5 times faster than Sonnet 4.5
- Instruction Following: 65% accuracy versus 44% from premium tier models in slide generation tasks
- Cost Efficiency: Enables frontier-level performance at a fraction of previous costs
Real-World Applications
The model's balance of intelligence and speed opens new use cases that weren't economically viable with larger models. For example, Claude Sonnet 4.5 can break down complex problems into multi-step plans, then orchestrate multiple Haiku 4.5 instances to complete subtasks in parallel—creating efficient multi-agent workflows.
Customers report that the model is particularly effective for:
- Sub-agent orchestration: Managing multiple AI agents working on related tasks
- Computer use tasks: Automated interaction with software interfaces
- Development workflows: Real-time AI assistance that feels instantaneous
- High-volume applications: Processing large numbers of requests cost-effectively
Safety and Alignment
Anthropic conducted extensive safety and alignment evaluations on Claude Haiku 4.5. The model demonstrated low rates of concerning behaviors and showed substantial improvements over its predecessor, Claude Haiku 3.5.
Notably, in automated alignment assessments, Haiku 4.5 showed statistically significantly lower rates of misaligned behaviors than both Claude Sonnet 4.5 and Claude Opus 4.1—making it Anthropic's safest model by this metric.
Safety testing indicated that Claude Haiku 4.5 poses only limited risks regarding CBRN weapons production. Consequently, Anthropic released it under the AI Safety Level 2 (ASL-2) standard, which is less restrictive than the ASL-3 classification applied to Sonnet 4.5 and Opus 4.1.
Availability and Pricing
Claude Haiku 4.5 is available immediately across all Anthropic platforms:
- Claude API: Use model identifier claude-haiku-4-5
- Claude Code: Available to all users
- Claude Apps: Efficiency improvements allow more usage within existing limits
- Cloud Platforms: Amazon Bedrock and Google Cloud Vertex AI
Pricing: $1/$5 per million input and output tokens—Anthropic's most economical price point while maintaining premium model performance.
The model serves as a drop-in replacement for both Haiku 3.5 and Sonnet 4, making it easy for existing users to upgrade and immediately benefit from the performance and cost improvements.