The New Stack reports on a compressed reasoning model aimed at AI agents.
The New Stack posted about Multiverse Computing's Quasar 438B, described as a compressed 438-billion-parameter reasoning model built for AI agents. The accompanying article states that Multiverse claims the model runs fast enough for agents, yet notes that benchmark speed does not always produce fast agents in practice. The post links to coverage that presents the model's performance claims alongside a more complicated picture from the benchmarks.
700
1 post, first seen 3d ago
The New Stack reports on a compressed reasoning model aimed at AI agents.
The New Stack posted about Multiverse Computing's Quasar 438B, described as a compressed 438-billion-parameter reasoning model built for AI agents. The accompanying article states that Multiverse claims the model runs fast enough for agents, yet notes that benchmark speed does not always produce fast agents in practice. The post links to coverage that presents the model's performance claims alongside a more complicated picture from the benchmarks.
The New Stack@thenewstackMultiverse Computing's Quasar 438B is a compressed 438-billion-parameter reasoning model built for AI agents, but benchmark speed doesn't always mean fast agents.
The New Stack@thenewstackMultiverse Computing's Quasar 438B is a compressed 438-billion-parameter reasoning model built for AI agents, but benchmark speed doesn't always mean fast agents.
