Sign in
VERIFIEDNov 222026

Is REAX 1.3 state of the art?Close.Near-SOTA in Decision models, rank #3.

As of Nov 22, 2026, REAX 1.3 is near state of the art, rank #3 in Decision models (since Nov 22, 2026; sources Decision Index Full; methodology v1.0). Close in Decision models: rank #3, 0.9 percentile points behind the leader.

Computed automaticallyas of Nov 22, 2026methodology v1.0confidence: medium
View as of point-in-time crowns
At a glance

Facts and scores

REAX 1.3 is an open-weights (Apache 2.0 for the blended work, with upstream licenses and notices preserved (component weights Apache 2.0 or MIT); REAX Engine inference code MIT) text model from REAX; 57B parameters and no public price.

How it scores

  • On Decision Index Full it scores 72.6 (CI 72.1–73.1), #4 of 420 (top 1%); 1.3 behind karpo-decision-31B; kudhab (Gemma 4 31B) is 0.4 above and Logits Sigma-1 Duo is 0.3 below.
  • On Decision Index ECE it scores 1.8%, #13 of 420 (top 4%); 0.7 pts behind KnowLine 9B Gen2; crystal-0.8b is 0.1 pts above and level with Copernicus-31B below.
  • On Decision Index Brier it scores 24.8%, #14 of 419 (top 4%); 4.4 pts behind lg-decision-4B; Logits Sigma-1 Duo is 0.1 pts above and Jeffan-31b-1710 is 0.1 pts below.

History

First result on isitsota: Nov 21, 2026. Best overall rank so far: #1 in Decision models, first held Nov 6, 2026. Crown intervals: Decision models Nov 6, 2026–Nov 11, 2026 and Decision models Nov 11, 2026–Nov 12, 2026.

Calibration

Reliability

Index bins for REAX 1.3. Confidence on the horizontal axis, accuracy on the vertical axis.

Median 245.9 ms, same GPU.

Receipts

Evidence

Scores come from the latest ingested snapshot
SourceScoreDateTypeReceipt
Decision Index Full
REAX 1.3
72.6 72.1–73.12026-11-21INDEP.snap_decision-index_2026-11-21_release-v4|decision-index-0.4.1
Reign history

Where it sits in the succession

Full timeline →
Decision models
Perplexity Decider v1.1 27B (Qwen3.8-27B, noncausal decision readout)
MISA AI 2.0 Decision Plus · oak (31B)
Perplexity Decider v1.1 27B (Qwen3.8-27B, noncausal decision readout)
vLLM-SR d3 (27B)
Perplexity Decider v1.1 27B (Qwen3.8-27B, noncausal decision readout)
StartLux-Decision v1.1 27B
baselabs mammoth-1 (27B)
Perplexity Decider v1.1 27B (Qwen3.8-27B, noncausal decision readout)
Fastino GLiDE 1.2
Perplexity Decider v1.1 27B (Qwen3.8-27B, noncausal decision readout)
Fastino GLiDE 1.2
KnowLine 31B Gen1
Perplexity Decider v1.1 27B (Qwen3.8-27B, noncausal decision readout)
Sieve 31B v3
GLiDE 27B
Perplexity Decider v1.1 27B (Qwen3.8-27B, noncausal decision readout)
REAX 1.3
GLiDE 27B
karpo-decision-31B
Perplexity Decider v1.1 27B (Qwen3.8-27B, noncausal decision readout)
REAX 1.3
GLiDE 27B
karpo-decision-31B
Perplexity Decider v1.1 27B (Qwen3.8-27B, noncausal decision readout)
GLiDE 27B
karpo-decision-31B
kudhab (Gemma 4 31B)
Perplexity Decider v1.1 27B (Qwen3.8-27B, noncausal decision readout)
Honest about the gaps

Where it is not the crown

CategoryThis modelRankLeader
Decision modelsNear-SOTA#3karpo-decision-31B
Links

Related pages

Is REAX 1.3 SOTA?

As of Nov 22, 2026, REAX 1.3 is near state of the art, rank #3 in Decision models (since Nov 22, 2026; sources Decision Index Full; methodology v1.0).

What is REAX 1.3 best at?

As of Nov 22, 2026, REAX 1.3 is near state of the art, rank #3 in Decision models (since Nov 22, 2026; sources Decision Index Full).

How does REAX 1.3 rank in Decision models?

Best current rank: #3 in Decision models, as of Nov 22, 2026.

When was REAX 1.3 last SOTA?

As of Nov 22, 2026, REAX 1.3 last held Decision models from Nov 6, 2026 until Nov 11, 2026.