Hermes Agent has introduced MoA presets as virtual models that outperform standard public options, showing an 8% improvement over Opus 4.8 and an 11% increase compared to GPT 5.5 on an upcoming benchmark. This development comes amidst a landscape where the strongest AI models are gated and access is restricted to a select group of users, highlighting the growing focus on exclusive and advanced AI technology. A full leaderboard for the new HermesBench is expected to be released soon.

GPT: GPT refers to OpenAI’s frontier AI model series, with version 5.5 noted in recent performance comparisons. It represents one of the publicly available top-tier models used as a reference point for Hermes Agent’s new offerings.
Opus: Opus refers to a leading frontier AI model series, with version 4.8 highlighted in recent comparisons. It serves as a key public benchmark that Hermes Agent claims to surpass in specialized evaluations.
Hermes Agent: Hermes Agent is an AI platform focused on advanced model orchestration and agent capabilities. It recently introduced exposure of MoA presets as virtual models to deliver performance exceeding publicly available frontier systems. This update is positioned as part of their upcoming HermesBench evaluation.

Model Access: The strongest AI models remain gated with access limited to select users.
Benchmark Development: New benchmarks like HermesBench are being prepared to evaluate capabilities beyond standard public models.