Kimi K3.1 Pro Ranks 4th on Artificial Analysis Index, Closing Open-Weight Gap

On October 24, Artificial Analysis published its evaluation of Kimi K3.1 Pro, Moonshot's open-weight model: it scored 55 on the Intelligence Index v4.3.2, up 11 points from Kimi K3's 44, ranking fourth globally behind Claude Opus 5.5 (58) and Claude Sonnet 5.5 (56) while surpassing GPT-5.1 and others—all at a 33% lower price. The community (@YuchenjUW, @haider1, Reddit user @No-Head-Royal) broadly sees this as evidence that open-weight models have caught up with the closed-source top tier, and possibly the strongest open-weight model to date.

Confirmed

• Intelligence Index score of 55, up 11 points from Kimi K3 (44), ranking fourth (@ArtificialAnlys, @YuchenjUW)
• Major agent gains: GDPval-AA at 1,846 Elo for third place, up 312 Elo from Kimi K3; Terminal-Bench 4.0 record raised to 67% (@ArtificialAnlys)
• Identical parameter count to Kimi K3 (2.8T total, 104B activated), yet tops the Pareto frontier on both "intelligence vs total parameters" and "intelligence vs activated parameters" (@ArtificialAnlys)
• Improved knowledge and hallucination: AA-Omniscience index up from 20 to 30, accuracy up from 48% to 51%, hallucination rate down from 53% to 44% (@ArtificialAnlys)

Why it matters

• Gaining 11 points at the same parameter scale purely through training/improvements, plus a 33% price cut, shows open models rapidly closing in on—or even beating—some closed-source flagships on efficiency and value
• The large jumps on agent benchmarks (GDPval-AA, Terminal-Bench) signal markedly stronger open-model competitiveness in real-world agentic scenarios

2026-10-24 ~ 2026-10-24 · 11 related posts

Full story(7 episodes)→

• Episode 1: Leaks Point to Imminent Kimi K3.1 Launch with 3D Rendering and Low Cost(2026-10-22, 17 posts)
• Episode 2: Moonshot teases Kimi K3.1 after strong Arena showing(2026-10-23, 4 posts)
• Episode 3: Kimi K3.1 Pro Rolls Out: 3T-Parameter Flagship with 1M Context(2026-10-24, 12 posts)
• Episode 4: Moonshot Releases Kimi K3.1 Pro, Open-Weight Model Rivaling GPT-6(2026-10-24, 33 posts)
• Episode 5: Nathan Lambert Calls Kimi K3.1 Pro the Strongest Open-Weight Model(2026-10-24, 2 posts)
• Episode 6: Kimi K3.1 Pro Ranks 4th on Artificial Analysis Index, Closing Open-Weight Gap(2026-10-24, 11 posts)
• Episode 7: Artificial Analysis Publishes Full Kimi K3.1 Pro Benchmark Charts(2026-10-24, 2 posts)

• [source] Kimi K3.1 Pro debuts 4th on Intelligence Index with big agentic gains, prices cut 33% — ArtificialAnlys · 2026-10-24
• [source] Kimi K3.1 Pro posts 67% on Terminal-Bench 4.0, highest of any model — ArtificialAnlys · 2026-10-24
• Kimi K3.1 Pro's hallucination rate drops to 44%, lifting AA-Omniscience to 30 — ArtificialAnlys · 2026-10-24
• Kimi K3.1 Pro: 2.8T total, 104B active, top of the efficiency Pareto frontier — ArtificialAnlys · 2026-10-24
• [source] Kimi K3.1 Pro hits #4 on Artificial Analysis Intelligence Index, closing on frontier closed models — Yuchenj_UW · 2026-10-24
• Kimi K3.1 Pro ranks 4th on AA Intelligence Index, may be strongest open-weight model — haider1 · 2026-10-24
• Kimi K3.1 Pro Scores 55 on Artificial Analysis, Outranking GPT-5.1 and Gemini 4 — No-Head-Royal · 2026-10-24
• Kimi K3.1 Pro Debuts at #4 on AA Index, Set to Narrow Open-Weights Gap to 3 Points — dbreunig · 2026-10-24

3 near-duplicate retellings: ArtificialAnlys · AmbassadorOk934 · NousResearch