Follow ZDNET: Add us as a favorite source On Google.
ZDNET Highlights
- Anthropic launches its new Haiku 4.5 model.
- The smaller model is faster and more cost-effective.
- Haiku 4.5 is available for all Cloud.AI free plans.
While we commonly associate large language models (LLMs) with generic AI products such as chatbots, small language models (SLMs) have their own advantages, including lower cost and higher speed. Leaning toward these advantages, Anthropic is now launching a new smaller model: Haiku 4.5.
Also: I Tested Free vs. Paid AI Coding Tools – Here’s Which One I’d Actually Use
Despite its smaller size, Haiku 4.5 demonstrated the same level of coding performance on the verified SWE-bench for Cloud Sonnet 4, which was launched four months earlier, with Anthropic claiming it was the world’s best coding model. The company said Haiku 4.5 can even outperform Cloud Sonnet 4 in tasks like computer usage at one-third the cost and twice the speed.
What does this mean for all cloud users
According to the post, the advanced capabilities, coupled with high speed, make it particularly suitable for real-time, low-latency tasks such as those that can be handled by chatbots. All users will be able to test Haiku 4.5 as the model is available across cloud apps, including the Claude.ai chatbot.
Beyond coding capabilities, the model performed competitively against the company’s flagship LLM Cloud Sonnet 4.5, launched in late September, on a series of benchmarks including MMMU, which tests for visual reasoning; AIME 2025, which tests for high school level mathematics; and r2-bench, which evaluates an agentic tool for the user.
On the security front, Anthropic said Haiku 4.5 showed a lower rate of concerning behaviors and was “significantly more aligned” than its predecessor, Cloud Haiku 3.5, and “significantly more” aligned than Cloud Sonnet 4.5 and Cloud Opus 4.1, making it the company’s most secure model to date. Its performance in security assessments earned it the AI Security Level 2 standard, which is internal to Anthropic, not determined by a third-party tester or external organization. The company offered more details on the designation in the model’s system card.
Developer Features
In addition to enjoying a cheaper but still capable model, developers will also be able to use Haiku 4.5 in conjunction with the cloud-heavy Sonnet 4.5, which Anthropic claims is still the most powerful coding model available.
Also: Even the best AI agents fail with this protocol – what can be done
Sonnet 4.5 can be used to solve a problem and develop a multi-step plan that “a team of multiple Haiku 4.5s” can complete task by task, Anthropic wrote.
Want more stories about AI? Sign up for AI LeaderboardOur weekly newspaper.
The Cloud Haiku 4.5 API is available to developers on Amazon Bedrock and Google Cloud’s Vertex AI. Anthropic said pricing is now $1/$5 per million input and output tokens.
