Ant's Ling 3.1 Flash nearly doubles intelligence index to 41, with 1M context
ArtificialAnlys · x · 2026-10-06
Ant Group's AntLingAGI released Ling 3.1 Flash, a 560B-parameter (25B active) reasoning model with a 1M-token context window; open weights coming soon.
Key results
- Scores 41 on the Artificial Analysis Intelligence Index v4.3, up from 20 for Ling 3.0 Flash (August), with big agentic gains
- GDPval-AA v2 Elo 1,622 and AA-Briefcase Elo 1,400, rivaling GLM-5.3-Flash, Gemini 3.8 Flash (high), and DeepSeek V4.1 Flash (Max); AutomationBench-AA 62% and Terminal-Bench v4.0 33% (vs 3% and 0% before)
- AA-Omniscience jumps to +2 from -18: accuracy up 18%→29%, hallucination down 44%→38%, showing it knows more rather than abstains more
Pricing and efficiency
- $0.30/$0.90 per 1M input/output tokens, cached input $0.06
- $0.99 per task—above GLM-5.3-Flash ($0.42) and DeepSeek V4.1 Flash ($0.32), below Gemini 3.8 Flash High ($1.24)
- 16% fewer output tokens than its predecessor on the full benchmark
Available via Novita AI now, weights to follow.
Related event: Ant Group's Ling 3.1 Flash Doubles Intelligence Index to 41(4 posts)→
More from Models
- GPT-6.1 Sol debuts at #5 on PostTrainBench, behind GPT-6 Astra and Opus 5.5 — maksym_andr · 2026-10-06
- ChatGPT can't decide which language to title its chats in — cool101wool · 2026-10-06
- GLM 5.3 Flash runs locally on dual V100s: 93GB MoE weights at ~20 tokens/s — lxfater · 2026-10-06
- Qwen3-Next-80B on a 3090 with 16GB RAM: ~3x decode speedup via MoE expert substitution — Zestyclose_Reality15 · 2026-10-06
- Anthropic investigating elevated errors on Claude Opus 5.5 requests — ClaudeAI-mod-bot · 2026-10-06
- ARC-1: a 1.7B decision model answering in ~20ms on a 4060 Ti, free and local — KMatysek · 2026-10-06