Sep 24, 2026 20:22 UTC
lmarena
lmarena.ai/leaderboard
LMArena: 91 new models on the leaderboard (andwise-evfd-agent, contenders/andwise-evfd-agent, contenders/deep-octo-first-party…) — anonymous, likely a stealth model
LMArena: 91 new models appeared on the leaderboard (labs test unreleased models here — often weeks before launch) [anonymous codenames — likely stealth] (25) + andwise-evfd-agent + contenders/andwise-evfd-agent + contenders/deep-octo-first-party + contenders/farris-ees5-agent + contenders/hy3-tencent-cloud-agent + contenders/inkling-agent + contenders/inkling-small + contenders/kivine-wxzc-agent + contenders/mimo-v2.5-pro-agent + contenders/muse-spark-1-3-max-agent + contenders/muse-spark-1.2-xhigh-agent + contenders/paisley-n9x0 + contenders/phoenix-qcs2-agent + contenders/skimaki-kwc6-agent + deep-octo-first-party + farris-ees5-agent + hy3-tencent-cloud-agent + inkling-agent + kivine-wxzc-agent + mimo-v2.5-pro-agent + muse-spark-1-3-max-agent + muse-spark-1.2-xhigh-agent + paisley-n9x0 + phoenix-qcs2-agent + skimaki-kwc6-agent [named / preview] (66) + claude-fable-5-agent + claude-fable-5.1-max-agent + claude-opus-4-8-thinking-agent + claude-opus-5-high-agent + claude-opus-5-max-agent + claude-sonnet-4-6-agent + claude-sonnet-5-agent + contenders/claude-fable-5-agent + contenders/claude-fable-5.1-max-agent + contenders/claude-opus-4-8-thinking-agent + contenders/claude-opus-5-high-agent + contenders/claude-opus-5-max-agent + contenders/claude-sonnet-4-6-agent + contenders/claude-sonnet-5-agent + contenders/deepseek-v4-flash-high-20260730-agent + contenders/deepseek-v4-pro-high-20260813-agent + contenders/deepseek-v4-pro-thinking-first-party + contenders/deepseek-v4.1-flash-max-20260910-agent + contenders/gemini-3.1-pro-preview-agent + contenders/gemini-3.5-flash-lite-agent + contenders/gemini-3.6-flash-agent + contenders/glm-5.2-agent + contenders/glm-5.3-agent + contenders/glm-5.3-flash-agent + contenders/gpt-5.4-high-code-arena-harness-agent + contenders/gpt-5.5-agent + contenders/gpt-5.5-xhigh-agent + contenders/gpt-5.6-luna-xhigh-agent + contenders/gpt
caught by the 20-minute sweep · the live wire · all receipts