Google DeepMind released Gemini 4 Neon, which lands at #2 overall on the Agent Arena Index with 1336 points and sits on the cost-performance Pareto frontier. The ranking draws on 4,200 real-world agentic sessions, surpassing Carbon, Hydrogen and Argon, with reviewers saying it comes close to the frontier.
Pareto Frontier
帕累托前沿
A set of optimal solutions where improving one objective (e.g., performance) requires sacrificing another (e.g., cost)
Curated Intelligence
Gemini 4 Neon (High) scored 1336 on the Agent Arena Index, ranking #2 overall and becoming Google DeepMind's top model on the leaderboard while joining the Pareto frontier. At a median cost of $2.82 per task, it undercuts #3 Claude Opus 5.5 (Max) by 53% and #1 Claude Fable 5.2 (Max) by 69%. It also ranks in the top three on praise-vs-complaint, steerability, and task completion.
OpenAI has launched GPT-6.2 Astra and GPT-6.2 Sol, with Astra targeting demanding science, legal and finance workloads and Sol optimized for fast everyday tasks, while cutting API pricing by 20% versus its predecessor. Astra ranks second on the Artificial Analysis Intelligence Index behind Claude Fable 5.2 but dominates the cost-performance Pareto frontier.
Microsoft AI's MAI-Image-2.7 has exited preview and now sits on the Pareto frontier across all three Arena.ai image leaderboards at $0.035 per image. It ranks #3 on Multi-Image Edit, #4 on Single Image Edit and #4 on Text-to-Image, gaining roughly 50 points over version 2.6, with photorealistic imagery rising from #8 to #3 and text rendering from #13 to #4.
Wincon Research had 16 frontier and open-weight LLMs attempt all 47 TIS-100 assembly puzzles with the community leaderboard withheld; the models produced 133 new frontier points, new records on 18 puzzles, and 3 outright cycle records. Human players then picked up and improved most of the generated programs within days, showing the outputs were genuinely usable engineering artifacts.
Arena.ai's weekly update reports that Moonshot released open weights for Kimi K3.1 Pro, which tops the Agent Arena open-model leaderboard with a 5.3% net gain. Tencent Hunyuan's Hy4 became the #1 Chinese model in Agent Arena at a $0.57 median cost per task and reshaped the Pareto frontier, with open weights promised soon. Google's Gemini Omni 1.2 Flash debuted at #1 in both text-to-video and image-to-video, while OpenAI's GPT Image 2.6 variants arrived at roughly twice the speed of their predecessor.
Tencent released Hunyuan Hy4, its latest flagship LLM, adding vision understanding for the first time and substantially scaling pretraining and post-training. First-day demand overwhelmed capacity, prompting a tenfold expansion of the WorkBuddy inference cluster and higher free quotas. Hy4 ranks second among Chinese models on Artificial Analysis' intelligence index and first among open models on Agent Arena, at roughly $0.78 per task.
Tencent Hunyuan's Hy4 climbed to #14 overall on Agent Arena across 12K+ real-world long-horizon agentic sessions, making it the top-ranked Chinese model and the only Chinese lab in the top five. It shows a +5.4% net improvement at a $0.57 median cost per task, scoring within 1 point of Kimi K3.1 Pro at roughly 62% lower cost.
After Arena.ai revised the Bash Recovery signal in its Agent Arena, Xiaomi's MiMo V2.6 Pro took the No. 1 open-model spot with a +3.8% net improvement, with the Flash variant at #4. The two are the only open models on the Arena Pareto frontier: Pro costs $0.11 per task, 15% less than #2 DeepSeek V4.1 Flash and 71% less than #3 Hy4 preview.
Tencent Hunyuan's Hy4 ranked #14 overall on Agent Arena across 8,500+ real-world long-horizon agentic sessions, up 3.6 points from its preview and joining the Pareto frontier. At $0.55 per task it costs roughly 62% less than Kimi K3.1 Pro Max ($1.46) while scoring within 1 point. Tencent says Hy4 will tie Kimi K3.1 Pro for #1 among open models once weights are released; it is currently proprietary.