Leaderboards miss vision skills: why DeepSeek's chart-reading progress doesn't show up in rankings
TwoBlueFools · reddit · 2026-11-13
The author questions relying on Artificial Analysis leaderboards to pick models for real work: actual usage involves screenshots, charts, dashboards and documents, while the Intelligence Index remains mostly text-based and doesn't reveal how well a model interprets complex charts.
DeepSeek added native vision with v4.1 Flash in September, and Tencent's Hy4 gained visual understanding in November — but neither is meaningfully reflected in overall rankings.
Testing Hy4 on financial charts, the author found it handled overall trends well and could turn analysis into a research brief, though numbers still need double-checking. Conclusion: overall scores may hide how good or bad models are at everyday tasks; test on your own work instead.
More from Models
- Claude back online after brief outage, developer confirms service restored — Salmaaboukarr · 2026-11-13
- Google appears to be dropping explicit caching: new API and Gemini 4 lack support — danson729 · 2026-11-13
- One prompt, 2,200 lines: ERNIE 5.5 Flash Preview builds animated sneaker site in under 5 min — Div_pradeep · 2026-11-13
- OpenAI's 1147-page Yang-Mills proof faces its first expert assessment: not proven yet — mishig25 · 2026-11-13
- Terminal-Bench 4.0 Cost Math: GPT-6.1 Sol Matches Opus 5.5 at ~1/8 the Price — LeadershipNo2904 · 2026-11-13
- HF staffer spotlights basal-2.0-mini, a 1.5B decision-tuned model from Bielik — lhoestq · 2026-11-13