Multiverse Computing 推出了其首个大规模语言模型 Quasar 438B1。该模型在 Artificial Analysis Intelligence Index v4.1.1 上获得 43 分，成为欧洲得分最高的 AI 模型1。在性能对标中，Quasar 438B 超越了 Mistral Medium 3.5（30 分）、NVIDIA Nemotron 3 Ultra（38 分）和 Inkling（42 分），仅次于 Claude Opus 5（63 分）1。
该模型在推理速度上表现突出，500 token 的输出耗时 15.3 秒，仅有 3 个同等规模模型速度更快1。在长上下文推理能力测试中，Quasar 438B 的 AA-LCR 得分达到 75.0，与 Grok 4.6 持平1；Terminal-Bench v2.1 测试得分为 69.3，领先 Mistral Medium 3.5 达 18.7 分1。该模型支持英语和西班牙语1，可通过 CompactifAI API 访问1。
Multiverse Computing has unveiled Quasar 438B, its inaugural large-scale language model and flagship inference model 1. The model achieved a score of 43 on the Artificial Analysis Intelligence Index, marking the highest performance among European AI models 1. Supporting both English and Spanish, Quasar 438B demonstrates considerable speed capabilities, generating 500 tokens in 15.3 seconds—a pace exceeded by only three competing models 1.
In comparative benchmarks, Quasar 438B outperforms several established models, including Mistral Medium 3.5 (scoring 30), NVIDIA Nemotron 3 Ultra (38), and Inkling (42), though it remains below Claude Opus 5 (63) 1. The model achieved a long-context reasoning score of 75.0 on AA-LCR, matching Grok 4.6's performance, and scored 69.3 on Terminal-Bench v2.1, exceeding Mistral Medium 3.5 by 18.7 points 1. Users can access Quasar 438B through the CompactifAI API 1.
评论
还没有评论，欢迎留下第一条。