Meta's Muse Spark 1.1 Benchmarks Comparable to Claude Opus 4.8
The release positions Meta's in-house model as a direct competitor to leading frontier models from Anthropic and OpenAI, with immediate availability through Meta's consumer and developer platforms.
Reporting from 1 source: GIGAZINE.
Meta Superintelligence Labs announced Muse Spark 1.1, an upgrade to the multimodal reasoning model released in April 2026. The model handles a 1-million-token context window, improves coding performance on large codebases, and scores on par with Claude Opus 4.8 in Meta's internal benchmarks, surpassing GPT-5.5 in several tests. It is available in the Meta AI app and via the Meta Model API.
Meta Superintelligence Labs has released Muse Spark 1.1, the follow-up to the multimodal reasoning model that debuted in April 2026. The new version is already in daily use by developers and researchers across Meta's entire organization, according to the company. In Meta's proprietary coding evaluation, Muse Spark 1.1 outperforms GPT-5.5 and shows performance comparable to Claude Opus 4.8, recording top scores in multiple benchmark tests. The model can actively manage a 1-million-token context window, retain information across long sessions, and operate external apps and services on behalf of users. It is available in the Thinking mode of the Meta AI app and at meta.ai, and can also be accessed for development through the Meta Model API.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.