Kimi K3 posts top AA-Briefcase results, but costs $10.57 per task and 56.4 minutes
ArtificialAnlys · x · 2026-07-22
Kimi K3 shows very strong frontier performance on AA-Briefcase, but at a heavy runtime and cost penalty.
- Average time per task is 56.4 minutes, among the highest recorded on the benchmark.
- The long wall-clock time is driven by 83 turns per task and about 120k output tokens per task, both well above Kimi K2.6.
- The model's cost per task is $10.57, one of the highest on the leaderboard.
- Compared with rivals, GPT-5.6 Sol (max) is 42 Elo lower while costing about 50% less, and Grok 4.5 (high) is 220 Elo lower while costing about 90% less.
Related event: Kimi K3 Ranks Second on AA-Briefcase but with High Costs and Long Runtimes(6 posts)→
More from Infra
- Musk: robot RAM demand will dwarf everything else, and 100% of superintelligence compute goes to deep space — elonmusk · 2026-10-27
- RTX 5090 hits $7K as PC gaming community goes into 'denial' over AI boom — PAstynome · 2026-10-27
- Wan2GP lands on Pinokio: 1-click AI video on 6GB GPUs — cocktailpeanut · 2026-10-27
- AWS on agent prompt caching: explicit + implicit patterns to slash token costs — AWS ML Blog · 2026-10-27
- Thieves spoofing GPS steal $100M+ in data center cargo since June — Ars Technica AI · 2026-10-27
- Lam Research: weekly token usage up ~250x in 18 months — Beth_Kindig · 2026-10-27