Grok 4.5 Evaluated on Cost-Performance
rohanpaul_ai · x · 2026-07-18
This post uses Artificial Analysis's Intelligence Index task cost to evaluate the cost-performance ratio of Grok 4.5.
The core conclusions are:
- The per-task cost for Grok 4.5 is about $0.31
- It is roughly 9 times cheaper than Claude Fable 5
- It is roughly 6 times cheaper than Claude Opus 4.8
- It is roughly 3 times cheaper than GPT-5.6 Sol or Kimi K3
The post also mentions token consumption data: Grok 4.5 uses nearly 6 times fewer tokens than Fable 5 on FrontierSWE, while still maintaining a top spot on the leaderboard.
The author argues that for many real-world applications, "cost per completed task" is more important than just edging out benchmarks, because it directly impacts:\n- agent cost\n- usage limits\n- testing speed\n- profit margins at scale
Therefore, the metric for frontier model competition should shift from "raw intelligence" to "efficient intelligence".
Related event: Grok 4.5 draws attention for low-cost, high-efficiency performance(10 posts)→
More from Models
- DeepSeek cut KV cache per token from 389KB to 890B in two years as open-closed gap debate rages — teortaxesTex · 2026-10-27
- GPT-6 Astra controls dexterous hands zero-shot at just 5.2% success in 500 trials — stingning · 2026-10-27
- Unverified: fable 5.5 speculated at 4-5T parameters to outscale 1-3T Opus — haider1 · 2026-10-27
- Kimi K3 Is 'Open' Until You Succeed: License Caps Free Use at $20M Revenue — alex_verem · 2026-10-27
- Screenshot shows 'GPT-6 Astra Ultra' in ChatGPT model picker — unverified leak or joke — justalexoki · 2026-10-27
- DeepSeek Safety Eval Finds "Over-Attribution": Model Mistakes Ordinary Tasks for Tests — Much_Boysenberry9124 · 2026-10-27