Grok 4.6 Ties for Top Spot on Agentic Index with 4x Cost Efficiency
elonmusk · x · 2026-08-18
Grok 4.6 ties with Claude Opus 5 Max for the top spot on the Artificial Analysis Agentic Index. It achieves this by completing tasks in approximately 53 turns and 0.5B input tokens on average, compared to 103 turns and 2.0B tokens for Claude. With a cost of $0.84 per task, Grok's superior turn efficiency places it on the Pareto frontier of intelligence versus cost, making long-running workflows affordable at scale.
More from coding & agent
- Dev moves app UI into Claude and ChatGPT: 'software distribution is changing fast' — fekdaoui · 2026-10-03
- Paper: LLM swarms fail at 'recursive social improvement' — copying peers hurts exploration — maxhkw · 2026-10-03
- Agentic traders get templates, TWAP and Bracket orders in weekly Grok Bot feature drop — MurrLincoln · 2026-10-03
- Stas Bekman releases AI skill that audits cluster compute, network and storage in hours — StasBekman · 2026-10-03
- Simple agent split: Claude for creative work, Codex for computer and browser tasks — VibeMarketer_ · 2026-10-03
- Intent adds stacks UI, per-agent token breakdowns for multi-agent orchestration — LukeW · 2026-10-03