Home / Technology / Intercom's AI Outperforms Giants, But Hides Its Base
Intercom's AI Outperforms Giants, But Hides Its Base
26 Mar
Summary
- Intercom's new AI model, Fin Apex 1.0, surpasses GPT-5.4 and Claude Opus 4.5.
- The model achieves a 73.1% resolution rate, significantly better than competitors.
- Despite its performance, Intercom will not disclose the base model used for Apex.
Intercom, a 15-year-old customer service platform, has announced Fin Apex 1.0, a custom-built AI model. This model reportedly outperforms leading frontier models such as OpenAI's GPT-5.4 and Anthropic's Claude Opus 4.5 in customer support metrics.
Fin Apex 1.0 boasts a 73.1% resolution rate, exceeding competitors by a notable margin. It also offers improved response times and a significant reduction in hallucinations. Intercom claims it operates at approximately one-fifth the cost of using frontier models directly.
Despite its claims of transparency, Intercom has declined to specify the base model or parameter size of Apex 1.0, stating it's "in the size of hundreds of millions of parameters." The company emphasizes that the "frontier" is now in post-training, utilizing proprietary data.
The AI-first pivot has proven successful for Intercom, with its Fin AI agent nearing $100 million in annual recurring revenue. This strategic shift has led to substantial company growth and is positioning Intercom as a leader in specialized AI applications.
Fin Apex 1.0 is exclusively available through Intercom's Fin agent and is not offered as a standalone model or via external API. Intercom plans to expand Fin's capabilities into sales and marketing functions.
