Knowledge work
AA Intelligence Index
Artificial Analysis nine-eval composite: GDPval-AA v2, Tau3-Banking, Terminal-Bench v2.1, SciCode, HLE, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR.
As of 13 August 2026. Conditional pack. Artificial Analysis Intelligence Index v4.1.1, observed 13 August 2026. Nine-eval composite. Not a diligence winner.
The Index mixes GDPval-AA v2, tau-cubed Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, and AA-LCR. Those are different estimands. The composite is shown as AA publishes it, not as an Antidote ranking device.
Fable 5 on the live page is Adaptive Reasoning, Max Effort, with Opus 4.8 fallback. GPT-5.6 Sol, Terra, and Luna are max effort. Cost per Index task is a different estimand from Index points. Launch articles in June and July 2026 reported different cells on an earlier Index revision.
- Scale
- AA Intelligence Index v4.1.1, observed 13 August 2026
- Provenance
- Artificial Analysis model pages for Claude Fable 5 and GPT-5.6 Sol, Terra, and Luna, observed 13 August 2026. Third-party harness, not an Antidote run.
- Contamination
- High. Public AA harness. Index revision changes historical cells.
- Cannot establish
- A diligence-workflow pass rate, a product moat, or NPV. The composite is not used as a winner ranking here.
Metrics, kept separate
- Intelligence Index
- USD per Index task
- Underlying eval scores, kept separate
Third-party leaderboard · not an Antidote run · not a product score
Artificial Analysis model pages, observed 13 August 2026. Index points shown as a fraction of 100. Fable 5 is the Adaptive Reasoning, Max Effort, Opus 4.8 fallback configuration. GPT-5.6 rows are max effort. Cost is USD per Intelligence Index task, a different estimand. This nine-eval composite is not a diligence winner. July 2026 launch articles reported different cells on an earlier Index revision.
| System | AA Intelligence Index | USD per task |
|---|---|---|
| Claude Fable 5 (fallback) | 62% | $3.14 |
| GPT-5.6 Sol max | 61% | $1.23 |
| GPT-5.6 Terra max | 57% | $0.51 |
| GPT-5.6 Luna max | 52% | $0.05 |
What the figures bound
On the observed 13 August 2026 model pages, Fable 5 (fallback) is 62 Index points at $3.14 per task. GPT-5.6 Sol (max) is 61 at $1.23. Terra (max) is 57 at $0.51. Luna (max) is 52 at $0.05. Sol is close to Fable 5 on this composite at lower list cost per task. That is not TAT-QA, not ParseBench, and not a product score.
Index revision is why June 2026 Fable 5 launch coverage said 64.9 and the July GPT-5.6 article said Sol 59. Dated cells are not mixed into one unofficial ladder.
What this page does not conclude
- A diligence-workflow or vendor-product pass rate
- NPV, adoption, or seat compression from Index points
- Official TAT-QA, CUAD, or SWE-bench Verified
- A winner ranking across unlike underlying evals
Cited as calibration on these memos — not a score of those products.
Not attached to a software, advisory, or automation memo yet.