DeepSeek V4.1 Flash is the API sweet spot; Kimi K3.1 Pro tops open models at 46
drdanielbender · x · 2026-11-13
The author wraps up an open-weight model tier list by the Artificial Analysis intelligence index:
- DeepSeek V4.1 Flash: 33 points, 552B total / 16B active. Called the API sweet spot — extremely high usage on OpenRouter, capable enough for real work with very competitive token pricing, "where intelligence meets a usable budget".
- Kimi K3.1 Pro: the clear intelligence leader among open-weight models at 46 points; 2.8T total / 104B active, also the biggest — and the priciest for both hosting and API.
Together with the earlier picks (Qwen3.8-27B local, Flash-Next on consumer GPUs), it forms a complete local-to-API open-model buying guide.
Related event: Open-Weight Model Tiers: Kimi K3.1 Pro Leads at 46, Budget Picks for APIs(2 posts)→
More from Models
- Ghibli SVG test: Opus 5.5 took 1 hour and $18, Gemini 4 Hydrogen 15 min and $0.90 — iamfakhrealam · 2026-11-13
- 18 hours after launch, users are building 3D worlds and games with Gemini 4 Hydrogen — iamfakhrealam · 2026-11-13
- Users complain Gemini's new 'Auto' mode dumbs down answers by deciding thinking depth for you — StudyHard30 · 2026-11-13
- 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