Compare two models
Pick any two decision models to see where each ranks, which benchmarks and test items split them, and what each costs. The link updates as you pick, so you can share a comparison.
d1 Ultra ranks higher overall and finishes ahead in 98% of 1,000 reruns on redrawn test sets: a settled gap. Drex Max scores higher on 3 of 5 benchmarks, d1 Ultra on 2. How ranking works.
Benchmark by benchmark
Each benchmark's headline score and rank. Bars show the score relative to the field: 100 for the best model on that benchmark, 0 for the worst.
Test by test
Every puzzle, ticket and case both models answered, each scored on its benchmark's own measure. Most are level; the coloured ones are where they part ways. Hover one for both scores. Sudoku squares open that race's replay.
Cost and speed
Every model gets the same requests, so these compare like for like. Cost covers one run of all five benchmarks; speed is the median time to answer one request.
Cost, all benchmarks
d1 Ultra is 2.7× cheaper.
Time per request
d1 Ultra is 1.8× faster.
Among every ranked model
Table view
| Model | Overall score | Cost to run all five benchmarks | Median time per request |
|---|---|---|---|
| d1 Ultrabest value | 90 | $0.0881 | 0.29s |
| Drex Max | 84 | $0.241 | 0.51s |
| Decider v1.2best value | 76 | $0.0593 | 0.39s |
| Jev 1.14 | 69 | $0.148 | 0.25s |
| Clef | 63 | $0.251 | 0.38s |
| Span-02 | 62 | $0.0934 | 2.22s |
| Mercury Decide | 61 | $0.0887 | 0.45s |
| Decision-1 | 59 | $0.0965 | 0.40s |
| GPT-6 Luna Decisions | 58 | $0.446 | 0.31s |
| Drex v2.0 | 58 | $0.148 | 0.52s |
| Decider Flash v1.2 | 57 | $0.118 | 0.39s |
| d1best value | 51 | $0.0437 | 0.30s |
| Tev1 4B Experimental | 47 | $0.605 | 0.58s |
| Clef Omni | 46 | $0.607 | 0.30s |
| Kev 4B | 41 | $0.101 | 0.57s |
| Clef Flash | 41 | $0.122 | 0.39s |
| Solar Decide Flash | 27 | $0.989 | 0.98s |
| Classify-1 | 6 | $0.792 | 0.53s |