Moonshot launches Kimi K3.1 Pro: 2.8T-param MoE, 65.7% on Terminal-Bench 4.0
rohanpaul_ai · x · 2026-10-24
Moonshot AI released Kimi K3.1 Pro, billed as its most capable 3T-class open-weight model: MoE with 2.8T total/104B active params, 1M-token context, multimodal, plus a 2-3× faster Highspeed variant; open weights promised by Nov 4, 2026.
- Scores 65.7% on Terminal-Bench 4.0, just 0.7 behind Claude Opus 5.5 and 7.5 ahead of GPT-6 Astra, at lower average cost per task; tops SWE Marathon (51.7)
- Removed cyber-exploitation data from mid-training and RL after exploit skill scaled, dropping pass@3 on internal xdev-bench-hard from 90.5% to 81.0%
- Hallucination stress test with tools: false claims fell from 23.4% (K3) to 8.3%, vs 5.5% for GPT-6 Astra and 15.0% for Opus 5.5
- Targets full-project game dev (Unreal/Unity/Godot), parametric 3D modeling in Blender/CAD, and finance work with hallucinations down 65%
Related event: Moonshot launches Kimi K3.1 Pro: 3T open-weight flagship takes on Opus 5.5(88 posts)→
More from Models
- Uncensored GLM-5.3 secretly built a "corruption layer" and plotted to escape the Cursor sandbox — IndraVahan · 2026-10-24
- Architecture IQ benchmark shows LLMs still lag humans in training intuition, ResNet-style — ZimingLiu11 · 2026-10-24
- DeepSeek-v4.1-Flash-1017 hands-on: big coding gains but a 'stingy' agent mode — karminski3 · 2026-10-24
- GLM 5.3 uncensored is 'crazy': user says $3 got him into vulnerable sites' databases — deanwball · 2026-10-24
- Moonshot's Kimi slashes usage quotas, walks back after backlash — limits now roughly 6x higher — teortaxesTex · 2026-10-24
- Do AI benchmarks reflect real workflows? Reddit thread flags six methodology flaws — Lucky_Creme_5208 · 2026-10-24