--- title: Multiverse says its 438B model is fast enough for AI... description: Multiverse Computing launched Quasar 438B, a compressed 438-billion-parameter reasoning model aimed at coding and enterprise agents, scoring 43 on Artificial... canonical: https://daily.dev/posts/multiverse-says-its-438b-model-is-fast-enough-for-ai-agents-the-benchmarks-tell-a-more-complicated--z03k5or4o twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Multiverse says its 438B model is fast enough for AI agents. The benchmarks tell a more complicated story. | daily.dev og:description: Multiverse Computing launched Quasar 438B, a compressed 438-billion-parameter reasoning model aimed at coding and enterprise agents, scoring 43 on Artificial... og:url: https://daily.dev/posts/multiverse-says-its-438b-model-is-fast-enough-for-ai-agents-the-benchmarks-tell-a-more-complicated--z03k5or4o og:image: https://api.daily.dev/og/posts/Z03K5or4o.png og:image:alt: Multiverse says its 438B model is fast enough for AI agents. The benchmarks tell a more complicated story. og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Multiverse says its 438B model is fast enough for AI agents. The benchmarks tell a more complicated story. **[The New Stack](https://daily.dev/sources/newstack)** · 4 min read · 0 upvotes · 0 comments ## Summary Multiverse Computing launched Quasar 438B, a compressed 438-billion-parameter reasoning model aimed at coding and enterprise agents, scoring 43 on Artificial Analysis' Intelligence Index and 69.3 on Terminal-Bench v2.1 with roughly 183 tokens per second output speed. The company claims it's the highest-scoring European model on the Intelligence Index, beating Mistral Medium 3.5 and NVIDIA Nemotron 3 Ultra, but trails frontier models like Claude Opus 5 (89.1 on Terminal-Bench). Multiverse hasn't disclosed the compression ratio, source model, or hardware requirements, and Quasar is proprietary, accessible only via API, raising questions about whether reported speed translates to real agent workloads that involve tool calls and growing context. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What benchmark scores does Multiverse's Quasar 438B model get on Artificial Analysis and Terminal-Bench? Quasar 438B scores 43 on Artificial Analysis' Intelligence Index and 69.3 on Terminal-Bench v2.1, with an output speed of about 183 tokens per second. That places it ahead of Mistral Medium 3.5 (30 on the Intelligence Index) and NVIDIA Nemotron 3 Ultra (38), but well behind Claude Opus 5, which scores 89.1 on Terminal-Bench v2.1. _daily.dev helps engineers compare emerging agent models against frontier benchmarks before committing to one._ ### How much did Multiverse compress the Quasar 438B model and what hardware does it need? Multiverse has not disclosed the compression ratio applied to Quasar 438B or which base model it started from, despite claiming its CompactifAI technology can typically shrink models by 80% to 95% with minimal accuracy loss. The company also hasn't specified hardware or compute requirements, and the model is proprietary, accessible only through its API with no way to inspect weights. _track disclosure gaps like this on daily.dev when vetting proprietary models for production agents._ ### How fast does Quasar 438B respond compared to how long an AI agent task actually takes? Quasar starts responding in about 1.1 seconds and generates a 500-token response, including reasoning, in around 15.3 seconds, according to Artificial Analysis measurements. Those per-call numbers are fast, but a full agent task also involves waiting on tools, processing growing context, and making repeated model calls, so raw response speed doesn't guarantee fast end-to-end agent performance. _developers weighing model speed against real agent latency can follow this tradeoff on daily.dev._ --- Tags: [#ai](https://daily.dev/tags/ai), [#machine-learning](https://daily.dev/tags/machine-learning), [#ai-agents](https://daily.dev/tags/ai-agents) [View this post on daily.dev](https://daily.dev/posts/multiverse-says-its-438b-model-is-fast-enough-for-ai-agents-the-benchmarks-tell-a-more-complicated--z03k5or4o) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Multiverse says its 438B model is fast enough for AI agents. The benchmarks tell a more complicated story.","url":"https://daily.dev/posts/multiverse-says-its-438b-model-is-fast-enough-for-ai-agents-the-benchmarks-tell-a-more-complicated--z03k5or4o","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/multiverse-says-its-438b-model-is-fast-enough-for-ai-agents-the-benchmarks-tell-a-more-complicated--z03k5or4o"},"datePublished":"2026-09-02T20:22:36.123Z","dateModified":"2026-09-02T22:15:10.091Z","description":"Multiverse Computing launched Quasar 438B, a compressed 438-billion-parameter reasoning model aimed at coding and enterprise agents, scoring 43 on Artificial...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/ac55e65090a65643ce598f7b2f164aa8?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/ac55e65090a65643ce598f7b2f164aa8?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"The New Stack","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"The New Stack","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/newstack","url":"https://daily.dev/sources/newstack"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/multiverse-says-its-438b-model-is-fast-enough-for-ai-agents-the-benchmarks-tell-a-more-complicated--z03k5or4o","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai,machine-learning,ai-agents","timeRequired":"PT4M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"The New Stack","item":"https://daily.dev/sources/newstack"},{"@type":"ListItem","position":3,"name":"Multiverse says its 438B model is fast enough for AI agents. The benchmarks tell a more complicated story."}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/multiverse-says-its-438b-model-is-fast-enough-for-ai-agents-the-benchmarks-tell-a-more-complicated--z03k5or4o#faq","mainEntity":[{"@type":"Question","name":"What benchmark scores does Multiverse's Quasar 438B model get on Artificial Analysis and Terminal-Bench?","acceptedAnswer":{"@type":"Answer","text":"Quasar 438B scores 43 on Artificial Analysis' Intelligence Index and 69.3 on Terminal-Bench v2.1, with an output speed of about 183 tokens per second. That places it ahead of Mistral Medium 3.5 (30 on the Intelligence Index) and NVIDIA Nemotron 3 Ultra (38), but well behind Claude Opus 5, which scores 89.1 on Terminal-Bench v2.1. daily.dev helps engineers compare emerging agent models against frontier benchmarks before committing to one."}},{"@type":"Question","name":"How much did Multiverse compress the Quasar 438B model and what hardware does it need?","acceptedAnswer":{"@type":"Answer","text":"Multiverse has not disclosed the compression ratio applied to Quasar 438B or which base model it started from, despite claiming its CompactifAI technology can typically shrink models by 80% to 95% with minimal accuracy loss. The company also hasn't specified hardware or compute requirements, and the model is proprietary, accessible only through its API with no way to inspect weights. track disclosure gaps like this on daily.dev when vetting proprietary models for production agents."}},{"@type":"Question","name":"How fast does Quasar 438B respond compared to how long an AI agent task actually takes?","acceptedAnswer":{"@type":"Answer","text":"Quasar starts responding in about 1.1 seconds and generates a 500-token response, including reasoning, in around 15.3 seconds, according to Artificial Analysis measurements. Those per-call numbers are fast, but a full agent task also involves waiting on tools, processing growing context, and making repeated model calls, so raw response speed doesn't guarantee fast end-to-end agent performance. developers weighing model speed against real agent latency can follow this tradeoff on daily.dev."}}]} ```