Artificial Analysis publishes full Decision Index results across six model leaderboards
ArtificialAnlys · x · 2026-11-14
Artificial Analysis has published the full breakdown of its Decision Index evaluations, comparing mainstream models across six leaderboards:
- AA-Paperclip: DeepMax V1.1 leads at 73, with SAIG Shap 2, GLIDE v1.2, Deeper v1.3 and o3Juno close behind in a dense 73–51 pack.
- EvalSafe-ONET-AA: the highest-scoring and tightest board — Jax V1.4, GLIDE v1.2 and Deeperkim 27B among many models clustered at 85–91.
- When2Call: SAIG Shap 1.2, Mammoth 27B, Elephant 9B and SAIG Shap 1.1 tie at the top (95–96), testing when models should actually invoke tools.
- AA-Omniscience-Decision v1.1: the most discriminating board, scores only 55–66 with Gemini Decision 1 first.
- HLE-MultiChoice v1.1: hardest task overall (35–49), Gemini Decision 1 tops it at 49.
- AA-Symmetry v1.1: output symmetry, Mammoth 27B leads at 88 while the tail drops to 56.
Overall, Gemini Decision 1 and the SAIG Shap family are the most consistent across boards, while HLE-MultiChoice and Omniscience still have no clear winner.
Related event: Artificial Analysis Debuts Decision Index; SubQ Snap 1.2 Ties for Lead(4 posts)→
More from Models
- ChatGPT Plus users hit with usage caps, excluded features as tier gets downgraded — fyn_world · 2026-11-14
- Decision models tested: training buys speed and cost, not reasoning back — someone_12443 · 2026-11-14
- Hands-on DeepSeek's $0.06/hr ASR API: loses to a locally fine-tuned 1.7B model — AGI Hunt · 2026-11-14
- LLMs playing AoE2 keep mistaking gold piles for berries, showing vision is unsolved — max_paperclips · 2026-11-14
- Claude MCP goes offline in rare outage, as blogger flags too many failures to be coincidence — AiBreakfast · 2026-11-14
- Claude Fable 5.2 builds an FPS game from one prompt as US engineer hires fall below half of 2022 peak — APPSO · 2026-11-14