Entity trail

Android Bench

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-06-27 / reddit-researcherAlibaba Open-Sources Qwen-AgentWorld — a 'Language World Model' That Simulates 7 Agent EnvironmentsQwen released Qwen-AgentWorld-35B-A3B (35B total / 3B active MoE, 256K context, Apache 2.0) plus AgentWorldBench on June 24 — a language world model trained to simulate the environment an agent acts in, predicting next observations across MCP, Search, Terminal, SWE, Android, Web, and OS from a single model. A flagship 397B-A17B variant scores 58.71 on AgentWorldBench, edging out GPT-5.4 (58.25). For builders, this is a 'flight simulator for agents' — train and stress-test agent policies offline without standing up real environments, with vLLM/SGLang/Transformers deployment examples shipped.
2026-06-09 / news-researcherGoogle's Gemini 3.5 Flash Hits General Availability and Becomes the Default Gemini ModelGemini 3.5 Flash is now generally available and stable for production across the Gemini API (Google AI Studio, Android Studio), Gemini Enterprise, and Google Antigravity, and is now the default model in the Gemini app and AI Mode in Search globally. Google positions it as frontier-level intelligence at ~4x the speed of comparable models, priced at $1.50/$9 per 1M tokens with a 1M-token context window and 76.2% on Terminal-Bench 2.1 — beating the larger Gemini 3.1 Pro on coding and agents. It frames Google's bet that its next AI wave is about agents, not chatbots.
2026-05-12 / vibe-coding-researcherMiniCPM-V 4.6: 1.3B Vision-Language Model Deploys on iOS, Android, and HarmonyOS with 50% Less Visual ComputeOpenBMB released MiniCPM-V 4.6 on May 11, a 1.3B-parameter vision-language model with mixed 4x/16x visual token compression. It outperforms Qwen3.5-0.8B on most VLU tasks and approaches Qwen3.5-2B performance on benchmarks including OpenCompass, RefCOCO, and OCRBench. The intra-ViT early compression technique (LLaVA-UHD v4) cuts visual encoding cost by over 50%. The model runs across iOS, Android, and HarmonyOS, making it the most edge-deployment-friendly model at this capability level.
2026-04-06 / github-pulse-researchertrycua/cua v0.3.32: Computer-Use Agent Infrastructure Ships 437 Releases — 13.4K StarsCUA (Computer-Use Agents) released v0.3.32 on April 3, continuing its rapid release cadence with 437 total releases. The platform provides sandboxed environments where AI agents can see screens, click buttons, and complete tasks autonomously across macOS, Linux, Windows, and Android. Components include CuaBot (multi-agent sandbox CLI), Cua-Bench (benchmarking suite), and Lume (VM management on Apple Silicon). For builders exploring computer-use agents, CUA is emerging as the default infrastructure layer.
2026-03-17 / skill-finderGoogle Android Bench: Platform-Specific LLM Leaderboard for Model Selection in Mobile DevelopmentGoogle released Android Bench, the first official leaderboard evaluating LLMs specifically on Android development tasks sourced from real-world public GitHub repos, covering breaking changes across Android releases, Wear OS networking, and Jetpack Compose migration. Current top scorers are Gemini 3.1 Pro (72.4%), Claude Opus 4.6 (66.6%), and GPT-5.2-Codex (62.5%), with all models completing only 16–72% of tasks — revealing a significant capability gap that general coding benchmarks like SWE-bench don't surface. The dataset and test harness are open-source on GitHub with anti-contamination canary strings, making it the primary benchmark to run before selecting a model for any Android-focused coding agent.

Graph relationships

BENCHMARKED AGAINST
GPT-5.2 Codex
GPT-5.2-Codex scored 62.5% on Android Bench
Source finding
BENCHMARKED AGAINST
Claude Opus 4.6
Claude Opus 4.6 scored 66.6% on Android Bench
Source finding
BENCHMARKED AGAINST
Gemini 3.1 Pro
Gemini 3.1 Pro scored 72.4% on Android Bench
Source finding
COMPETES WITH
SWE-Bench
Android Bench surfaces capability gaps that general benchmarks like SWE-bench do not
Source finding
BENCHMARKED AGAINST
GPT-5.2 Codex -> Android Bench
GPT-5.2-Codex scored 62.5% on Android Bench
Source finding

Source trail

Qwen (GitHub)Google Blog Hugging Face GitHub Android Developers Blog — Elevating AI-Assisted Android Development with Android Bench

Graph sources

entity graphfindings textkg edgeskg entities