Artificial Analysis' Intelligence Index: Sonnet 5.5 (max) ranks above GPT-6 Astra (max) and behind only Opus 5.5 (max) but has the highest token use of them all
Context & Ripple Effects
Artificial Analysis’ earlier Intelligence Index snapshots put Opus 5 at 63 and Fable 5 at 62, while GLM-5.3 reached 60 with max reasoning. Sonnet 5.5’s placement reshuffles that top tier, placing it ahead of GPT-6 Astra on this broad measure while still below Opus 5.5.
The comparison also sharpens a distinction visible in Artificial Analysis’ coverage: GPT-6 Astra’s coding-agent score of 67 was competitive with several leading models, whereas the Intelligence Index ranks general capability differently. Token use adds an operating-cost dimension to the headline ranking.
First-order effects
- Sonnet 5.5 becomes the higher-ranked alternative to GPT-6 Astra on Artificial Analysis’ Intelligence Index, while Opus 5.5 retains the leading position.
- Teams evaluating Sonnet 5.5 must weigh its near-leading score against the index’s finding that its max setting uses more tokens than the compared models.
Second-order effects
- GPT-6 Astra and other models near the top of the index face comparison on two axes rather than one: benchmark position and token consumption per task.
- Buyers using Artificial Analysis’ results gain a reason to separate broad intelligence rankings from workload-specific tests such as coding-agent evaluations.
Third-order effects
- If maximum-reasoning configurations continue to dominate capability tables while consuming more tokens, model selection will increasingly turn on cost per useful task rather than rank alone.
- Independent benchmark providers are becoming a more consequential layer in AI competition as their indexes expose different leaders across general, coding, and efficiency measures.
The trend: Frontier-model competition is shifting from a single benchmark race toward a capability-versus-inference-efficiency trade-off measured across specialized evaluations.