Physix Frontier badgePhysix Frontier
PublishChinese | EN

Physix Frontier · News Briefing Card (Hacker News · Sep 2, 2026)

Multiverse Unveils 438B Reasoning Model Quasar

KEY FACTS

  • Multiverse has released Quasar, a 438-billion-parameter reasoning model supporting English and Spanish.
  • Quasar scored 43 on the Intelligence Index, surpassing Mistral's score of 30.
  • The model generated 500 tokens including thinking time in 15.3 seconds, faster than most comparable models.
  • It achieved a score of 75.0 on the AA-LCR long-context test, matching Grok 4.6.
  • The model is available via the CompactifAI API, targeting enterprise agents and coding tasks.

KEY DATA

438BParameter Size
43Intelligence Index
15.3s500-token Response
75.0AA-LCR Score

PHYSIX OBSERVATION

European models have historically faced skepticism about keeping pace with frontier developments, but this release prioritizes speed, long context, and API integration as key selling points. This signals a shift in competition from benchmark chasing to deployable efficiency. For enterprise users, a model that fits into interactive workflows may hold more value than one with higher scores but slower performance.

Source: Hacker News report