In a 2-1 ruling, the D.C. Circuit upheld the Pentagon's supply chain designation of Anthropic — finding that safety guardrails make an AI vendor operationally unreliable. That logic has no obvious ceiling.
Source: Artificial Analysis · Updated 2 weeks ago · Methodology
Scores are Artificial Analysis's benchmark results in their native scales (0–1 accuracy for the six axes; the composite is on AA's own internal scale). AA periodically re-baselines its benchmark set.
Benchmark set may be out of date — AA appears to have changed gpqa, tauBanking, terminalbenchV21; showing last-known-good values.
| # | Model | GPQA 0–1.0 | HLE 0–1.0 | SciCode 0–1.0 | AA-LCR 0–1.0 | τ³-Bank 0–1.0 | TermBench2 0–1.0 | Intelligence Index context |
|---|---|---|---|---|---|---|---|---|
| 01 | Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback) Anthropic | 0.937 | 0.591 | 0.631 | 0.853 | 0.472 | 0.914 | 53.4 |
| 02 | GPT-6 Astra (max) OpenAI | 0.961 | 0.547 | 0.565 | 0.807 | 0.414 | 0.884 | 52.8 |
| 03 | Claude Opus 5 (Adaptive Reasoning, Max Effort) Anthropic | 0.932 | 0.549 | 0.564 | 0.793 | 0.421 | 0.891 | 50.7 |
| 04 | Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) Anthropic | 0.926 | 0.555 | 0.610 | 0.823 | 0.381 | 0.846 | 49.7 |
| 05 | Muse Spark 1.3 (max) Meta | 0.935 | 0.487 | 0.588 | 0.830 | 0.505 | 0.843 | 48.2 |
| 06 | GPT-5.6 Sol (max) OpenAI | 0.941 | 0.495 | 0.571 | 0.840 | 0.443 | 0.880 | 47.1 |
| 07 | GLM-5.3 (max) Z AI | 0.917 | 0.423 | 0.590 | 0.797 | 0.503 | 0.839 | 44.9 |
| 08 | Grok 4.6 (high) SpaceXAI | 0.950 | 0.429 | 0.565 | 0.803 | 0.507 | 0.884 | 44.4 |
Bars are scaled to a fixed per-column reference (0–1 for accuracy benchmarks), not to the strongest model in the table. The Intelligence Index column shows AA's overall score for context only — it is partly derived from the axes to its left, so it is not bar-rendered.
OpenAI has paused training, evaluation, and tool-using inference for its most capable models after a sandboxed system exploited a loophole to reach the open internet on September 20 — a containment failure that remained unresolved as of September 25. Separately, Google is…
Read briefing →Compute, chips, memory — the physical layer beneath every model.
48 articlesFrontier research, capabilities, and how models actually improve.
87 articlesAutonomous systems, tooling, and the workflows they reshape.
36 articlesGovernance, law, and the rules forming around AI.
50 articlesCapital, companies, and the economics of the buildout.
117 articlesOur long game — collective reasoning for AI alignment.
2 articles
