Arena (LMArena) · measurers · lmarena ☆ FOLLOW 42 topic(s) over 9 day(s) — architecture, evaluation, inference. on X ↗
CARD NO. 2026-10-05 Google launches Gemini 4 Argon with a 1 million token output limit architecture · 23 voicesNew models achieve top rankings across AI evaluation benchmarks evaluation · 15 voicesOpenAI releases GPT-6 models including GPT-6 Astra and GPT-6.1 Sol architecture · 15 voicesAnthropic introduces Claude Sonnet 5.5 with higher speed and lower cost architecture · 14 voicesDeepSeek pauses free V4.1-Flash promotion following high abuse inference · 4 voicesAdvances in reinforcement learning and updates to video generation benchmarks evaluation · 3 voices
CARD NO. 2026-10-04 Google launches Gemini 4 Argon model for complex software engineering tasks architecture · 24 voicesClaude Code introduces support for UI and behavior customization via plugins agents · 17 voicesAnthropic launches Sonnet 5.5 model family and extends thinking persistence architecture · 17 voicesGoogle launches Gemini 4 Argon and Anthropic releases Claude Sonnet 5.5 evaluation · 14 voicesGemini 4 Argon supports 1M output tokens while Claude Fable 5.5 rolls out inference · 9 voicesDeepSeek Harness desktop versions released alongside Agent Arena Pareto updates agents · 5 voicesOpenAI launches GPT-6.1 Sol and releases it on Agent Arena evaluation · 4 voices
CARD NO. 2026-10-03 Claude Sonnet 5.5 releases with 30% faster speed and isolated context thinking inference · 23 voicesGoogle releases Gemini 4 Argon with a 1M token output limit architecture · 14 voicesClaude Code adds support for custom mods and UI modifications agents · 14 voicesFireworks Research releases Ember-1 built on Kimi K3 training · 10 voicesPerformance evaluations of Claude 5.5 and Fable 5.5 reveal reduced hallucination rates evaluation · 8 voicesLlama.cpp adds local support for decision models through the /v1/systemone endpoint inference · 6 voices
CARD NO. 2026-10-02 Qwen3.8-27B multimodal decision model and Qwen-Image-2.1 released multimodal · 10 voicesOpenAI announces GPT-6.1 Sol offering high efficiency at lower cost architecture · 4 voices
CARD NO. 2026-10-01 OpenAI releases GPT-6.1 and patches vision bugs across GPT-6 models architecture · 21 voicesAnthropic Releases Claude Sonnet 5.5 and Claude Code Limit Updates architecture · 16 voicesGoogle launches Gemini 4 Argon and Gemini 3.8 Flash TTS audio models architecture · 14 voicesAnthropic releases Claude Sonnet 5.5 with improved speed and efficiency architecture · 14 voicesAnthropic expands model lineup with Claude Sonnet 5.5 and Haiku 5.5 evaluation · 12 voicesPlatforms release new models and ultrafast token generation speeds inference · 10 voicesOpenAI Launches GPT-6.1 Sol Model Focused on Cost Efficiency architecture · 4 voices
CARD NO. 2026-09-30 Grok 4.7 claims top ranking on the AA cyber index evaluation · 22 voicesAnthropic launches Claude Sonnet 5.5 with speed and efficiency gains architecture · 15 voicesAnthropic releases Claude Sonnet 5.5 with 30% faster speed architecture · 14 voicesAnthropic releases Claude Sonnet 5.5, expands Claude Code features, and updates pricing policy architecture · 14 voicesNew AI models and tools released for lightweight tier inference · 9 voicesDeepSeek Harness releases official TUI client alongside new agent benchmarks evaluation · 7 voices
CARD NO. 2026-09-29 GPT-6 Sol and GPT-6 Luna launched with API prices 50% lower inference · 17 voicesAnthropic launches Claude Sonnet 5.5 with higher speed and lower cost architecture · 14 voicesOpenAI resolves service disruption affecting Codex system_design · 10 voices
CARD NO. 2026-09-28 OpenAI releases GPT-6 Sol and GPT-6 Luna models architecture · 12 voicesOpenAI and xAI announce model lineups and benchmark improvements evaluation · 11 voicesCursor integrates Claude Opus 5.5 while Claude Marketplace expands third-party tools agents · 8 voices
CARD NO. 2026-09-27 OpenAI introduces GPT-6 Sol and GPT-6 Luna models architecture · 19 voicesGitHub Copilot introduces Autopilot and integrates GPT-6 family models agents · 12 voices