Gemini Nearly Absent from LM Arena Coding Top 15 as Claude and GPT Dominate
Full_Piano_3448 · reddit · 2026-10-13
A Reddit post highlights a striking shift on the LM Arena coding leaderboard: Claude and GPT models dominate while Google's lineup is nearly absent from the top 15, with only Gemini 4 Argon at #7. An attached Agent Arena capabilities chart (802,643 real sessions, 52 models) shows Claude Opus 5.5 (Max) leading with +17.58% net improvement, GPT 6 Sol and GPT 6.1 Sol at #2-3, four Claude models in the top six, Grok 4.7 at #14, and DeepSeek V4 Pro / V4.1 Flash at #15-16.
The author wonders whether Google has shifted priorities toward multimodal AI, search, agents, and ecosystem integration rather than competing for the top coding spot, or whether this is just a gap before its next release. Commenters debate whether Google is genuinely falling behind in coding.
More from Models
- Decision Index 0.3.2 ranks 148 open reproductions of TypeSafe Jev's Decision Model — capetorch · 2026-10-13
- OpenAI's mysterious 'Muse wonder' project spotted, looks like a big model — altryne · 2026-10-13
- 3D demo animations are benchmaxed for tweets, not real model quality signals — qtnx_ · 2026-10-13
- GPT-6 Astra at Max reasoning effort impresses in real test but torches usage limits, user says — CtrlAltDwayne · 2026-10-13
- Hermes Hits 68 on the Leaderboard With No Regression, Says Decider Models Have Huge Headroom — antoine_chaffin · 2026-10-13
- Blogger: Anthropic quietly built the best all-in-one agent chat experience — iruletheworldmo · 2026-10-13