Grok 4.7 Lands at #16 in Agent Arena With +3.96% Gain, But Costs 54% More Per Task
arena · x · 2026-09-27
Agent Arena reports Grok 4.7 (xHigh) debuts at #16 with a +3.96% net improvement score — but at a notably higher cost: $1.14 median per task vs. $0.74 for Grok 4.6 (High, +1.22%) and $0.43 for Grok 4.5 (+1.50%).
Signal breakdown:
- Confirmed Success: #6 (+10.41%)
- Bash Recovery: #12 (+5.98%)
- Praise vs Complaint: #16 (+4.93%)
- Steerability: #28 (-1.89%)
- Tool Hallucination: no issues (+0.36%)
xAI markets 4.7 as better than 4.6 at the same price and speed, but Arena data shows both price and capability moving up together — worth watching from a cost-effectiveness angle.
More from Models
- Codex power users say Astra burns $200 limits in days while Anthropic's cap is hard to hit — brandon_galang · 2026-09-27
- Claude Opus 5.5 codes a 15-second Audi RS 6 motion piece in React/Remotion with synthesized V8 soundtrack — EricBuess · 2026-09-27
- Allie Miller: Opus 5.5 Is Great but 'Ridiculously Verbose' — Compares It to GPT-6 Sol — alliekmiller · 2026-09-27
- AI still misses the last 5% of design polish, argue designers as models near top UI kits — zielinskiwoj · 2026-09-27
- MLX poll: all top 3 community picks are powered by MLX-VLM — andrejusb · 2026-09-27
- Google AI Pro Subscriber Reports Gemini's Pro Model Missing, Only Flash-Lite Left — BassNPedal · 2026-09-27