Ant's open-source Ling-3.0-flash-VL edges out GPT-5.4 on Image-to-WebDevArena
智东西 · wechat · 2026-09-09
Ant Group open-sourced its first natively multimodal model, Ling-3.0-flash-VL: a 124B-parameter MoE (5.5B active) with native image/text/video input, 256K context, and BF16/FP8 weights on HuggingFace and ModelScope.
- Visual feedback loop: the model re-reads page/environment state after acting and iterates—screenshot → code → browser render → compare → refine; the same mechanism powers GUI agent workflows across apps.
- Benchmarks: 42 on Artificial Analysis index (+4 vs text-only); 1441 on Image-to-WebDevArena vs GPT-5.4's 1440; OmniDocBench 91.35, WebVoyager 90.83—though still behind frontier closed models like Claude Fable 5.1 (66).
- Use cases: sports highlight auto-editing, real-time camera Q&A in the Ling Guang app, structured extraction across multi-page medical reports (explicitly not a doctor replacement).
Overseas devs already recreated a playable Flappy Bird page from a screenshot, calling it "a pretty solid release."
More from coding & agent
- Two engineers rebuilt a FedEx supplier's billing core in 3.5 months vs an estimated 7-8 — alex_verem · 2026-10-27
- GPT-6.1 Terra gains ~20 points on Terminal-Bench 4.0, rivalling GPT-6.1 Sol — ArtificialAnlys · 2026-10-27
- GPT-6.1 Terra scores 48 on AA Coding Agent Index at one-third the cost of GPT-5.6 Terra — ArtificialAnlys · 2026-10-27
- Shift-1 router cuts coding agent spend up to 55% while keeping SWE-bench accuracy — ypatil125 · 2026-10-27
- Non-coder vibe-codes a Godot JRPG with Claude Code in two weeks, music and art synced — ADHDavid · 2026-10-27
- StackChan robot performs a drama skit via Hermes agent and GPT-Luna-6 — alexcovo_eth · 2026-10-27