MiMo-V2-Flash shows strong benchmark performance in GPQA, Long Context

MiMo-V2-Flash, an open-source model, has demonstrated strong performance on several key benchmarks, including GPQA and Humanity's Last Exam. The model achieved 83.5% on GPQA and 20% on Humanity's Last Exam, with notable scores in long-context reasoning (64.3%) and SciCode (38.3%). These results, measured independently, position MiMo-V2-Flash as a competitive option in the open-source AI landscape. AI

IMPACT This research provides new data points for the performance of open-source models on key AI benchmarks.

RANK_REASON The cluster reports benchmark results for an open-source model, which falls under research. [lever_c_demoted from research: ic=1 ai=1.0]

Read on Mastodon — fosstodon.org →

• GPQA: A Graduate-Level Google-Proof Q&A Benchmark
• Humanity's Last Exam
• long-context reasoning
• MiMo-V2-Flash
• SciCode

• model release
• paper

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →