Claude 3.5 Haiku API has been launched, but its high pricing has been criticized by users
Recently, Anthropic announced that Claude 3.5 Haiku is now accessible via API. Claude 3.5 Haiku is available on Anthropic's API, Amazon Bedrock, and Google Cloud's Vertex AI. Alex Albert, a representative, added that the model has also updated its knowledge up to July 2024—making it the most current of all Claude models.
Initially, there was a lot of anticipation for its launch, but once users saw the price of Claude 3.5 Haiku, they began to complain that it was excessively expensive! The price is set at $1 per million tokens for input and $5 per million tokens for output. Despite the performance enhancements, the price of Haiku has risen to four times that of previous models, which explains why users are finding it hard to accept.
So what makes Claude 3.5 Haiku so special that it warrants such a high price?
Anthropic states that Claude 3.5 Haiku only offers a pure text model, with future plans for image input support. Benchmark data shows that the performance of 3.5 Haiku is close to that of Claude 3.5 Sonnet, and it outperforms all previous Claude models except for the latest 3.5 Sonnet in programming and agent tasks.
Let’s take a look at some use cases applicable for Haiku as listed by Anthropic:
- Code Completion: Claude 3.5 Haiku offers fast and accurate code suggestions to enhance development efficiency.
- Interactive Chatbots: Its improved conversational abilities and quick responses make it suitable for large-scale user interactions in chatbots.
- Data Extraction and Annotation: It efficiently processes and categorizes information, making it relevant for financial, medical, and research institutions dealing with unstructured data.
- Real-time Content Review: Enhanced reasoning capabilities provide reliable content audits to ensure safety on social platforms and in media organizations.
Why are users dissatisfied with the pricing of Claude 3.5 Haiku?
Firstly, it’s essential to recognize that Claude 3.5 Haiku competes with OpenAI’s GPT-4o Mini and Google’s Gemini 1.5 Flash, and the upcoming competition results may offer insights.
We’ve already discussed the pricing for Haiku; let’s examine the pricing for GPT-4o mini and Gemini 1.5 Flash:
- GPT-4o Mini: OpenAI’s GPT-4o Mini charges $0.15 per million tokens for input and $0.6 per million tokens for output.
- Gemini 1.5 Flash: For prompts with fewer than 128,000 tokens, Gemini 1.5 Flash charges $0.075 per million tokens for input and $0.3 per million tokens for output; if exceeding 128,000 tokens, it costs $0.15 per million tokens for input and $0.6 per million tokens for output.
Some users have also created comparison tables showcasing the performance of the three models.
In summary, Claude 3.5 Haiku underperformed compared to GPT-4o Mini on many benchmark tests, despite its higher cost; it showed only slight performance improvement over Gemini 1.5 Flash, while being priced significantly higher. This explains why some find the pricing outrageous.
For example, XXAI, which aggregates multiple AI tools, offers users access to 13 AI models for only $9.9 per month, presenting a very competitive price point. Therefore, it’s essential to understand that regardless of robust features, a product's pricing must align with user expectations to attract a wider user base.
Anthropic has provided reasons for the increased pricing of Claude 3.5 Haiku on X.
They argue that in final tests, Haiku surpassed many benchmark tests compared to their prior flagship model, Claude 3 Opus, while the cost is only a fraction of the latter. As a result, they have raised the price of Claude 3.5 Haiku to reflect its enhanced intelligence.
Nonetheless, not everyone is willing to accept this pricing.
Anticipating Anthropic's Response
In light of developers’various opinions on the pricing of Claude 3.5 Haiku, Anthropic's response is undoubtedly a topic worth following. As more businesses and developers increasingly focus on and utilize AI technologies, the market's sensitivity to pricing continues to grow, especially as competition heats up between major players like OpenAI and Google. Thus, Anthropic must thoroughly consider how to adjust its pricing strategy to better meet developers' needs and market expectations.