Anthropic’s fastest version, Claude 3.5 Haiku, available now

Anthropic has officially released its Claude 3.5 Haiku version to all users via the Claude chatbot on the web and mobile apps, such as seen by AI power users on X.

Previously limited to developers who accessed it via Anthropic’s API following its launch in October 2024, this small, fast model has attracted attention for its ability to outperform large brands in benchmarks while maintaining a competitive price.

According to a third-party benchmarking organization Production AnalysisClaude 3.5 Haiku “has a low latency compared to the average, taking 0.80s to receive the first signal (TTFT),” however “and slower than the average, with a speed of 65.1 signals per second.”

The release – which has not been officially announced – comes after major updates from Anthropic’s AI rivals OpenAI and Google, who will also be sending new models to their forums as the year ends. OpenAI versions of o1 and o1-mini and Gemini 2 by Google.

The question for Anthropic is whether customers will be happy enough with the performance of Claude 3.5 Haiku to subscribe to its Pro version – or continue to use it instead of other advanced and faster versions.

Claude 3.5 Haiku is available via the Claude Chatbot

As the fastest and most affordable model in Anthropic’s series, Claude 3.5 Haiku excels in real-world applications such as processing large datasets, analyzing financial statements, and extracting results from long-form data.

It has a window of 200,000 symbols – more than the window of 128,000 symbols on OpenAI’s GPT-4 and GPT-4o – allowing it to handle more information easily.

For the Claude chatbot, Haiku brings functionality that enhances its flexibility. Users can scan images and file attachments, which makes them useful for social media projects and workflows for integrating large data sets.

Haiku also includes and Claude ArtifactsThe interactive side panel was first introduced in June 2024. The artifacts provide a dedicated workspace to control and update AI-generated content in real time, including using full-featured software. In my test of Artifacts and Haiku this morning, it managed to write a playable version of Pong in less than a minute:

Despite its power, Haiku has limitations. It doesn’t support web browsing or image rendering, both of which are offered by competitors such as OpenAI’s GPT-4o and GPT-4.

Additionally, my brief test this morning showed that it failed the “Strawberry Test,” a user-generated challenge where the AI ​​must recognize all three R’s in the word strawberry.

Access and registration information

Claude 3.5 Haiku is freely available through the Claude chatbot, but users face a daily message limit based on server requirements.

For example, on the free session this morning when I tested it, I managed about 10 exchanges (20 messages total in and out) before I reached the Anthropic session, which resets daily.

To unlock more usage, users can subscribe to the Claude Pro plan, priced at $20 per month.

This subscription offers 5x the usage of the free tier, access to high-level updates, access to new content, and access to additional versions like Claude 3 Opus.

The pricing structure reflects OpenAI’s ChatGPT Plus subscription, providing more information for power users.

Performance and price

On the API, Claude 3.5 Haiku offers exceptional service at an affordable price. Starting at $0.80 per million entry tokens and $4 per million tokens, it offers an economical solution compared to large brands like Claude 3 Opus.

Developers can also reduce costs by using fast caching, which saves up to 90%, and the Message Batches API, which cuts costs by 50%.

In benchmark testing, Haiku has outperformed major, publicly available brands. Its performance includes a 40.6% score on the SWE-bench Verified, the main indicator of literature, showing its strength in tasks that require intelligence and speed. This makes Haiku an excellent choice for user-focused applications and long-term flooding.

Key points

Although Claude 3.5 Haiku offers powerful capabilities, users should consider its limitations. The lack of web browsing and image processing may make it less appealing in some use cases compared to its competitors. In addition, the daily message cap may be difficult for users who do not want to upgrade to a Claude Pro subscription.

However, with features such as image and file scanning, robust scrolling capabilities, and integration with Artifacts, Haiku remains a powerful tool for tasks that require speed and accuracy.

The Artifacts feature, in particular, extends its functionality beyond voice output, enabling interactive editing and virtual reality manipulation.

For users who are ready to explore their potential, Claude 3.5 Haiku is now live and available through the Claude chatbot online and mobile apps for iOS and Android.