Benchmarks › GDPval-AA
68.3top score · Claude Opus 5.5
175models ranked
106open-weights models
Crowd-BT Elohigher is better
Oct 20, 2026last updated
How the test works
1
Prompts
Public gold GDPval tasks
- What the model doesCreate files that satisfy real professional work requests
- Kinds of tasksProfessional work
- In → outText, reference files → Files
2
Votes
Pairwise comparisons fitted with Crowd-BT (v1.1 Briefcase / v2.1 GDPval)
3
Rating
Crowd-BT Elo
- Better isHigher
Sources: artificialanalysis.ai
GDPval-AA 2.1 leaderboardtop 10 of 175
Full leaderboard with filters →| Model | GDPval-AA 2.1 | Per task | Price | Speed | Supports | Context |
|---|---|---|---|---|---|---|
68.3max | $5.98 | $4 / $20 | 96 | 1M | ||
67.9 | — | — | 1M | |||
66.9max | $5.46 | $2 / $10 | 137 | 1M | ||
63.5max | $4.52 | $4 / $20 | 161 | 1M | ||
62.7max | $7.63 | $10 / $50 | 69 | 1M | ||
62.3max | $0.26 | $0.30 / $1.20 | 167 | 1M | ||
61.1max | $5.86 | $5 / $25 | 55 | 1M | ||
60.6xhigh | $3.74 | $2 / $6 | 94 | 500K | ||
59.2thinking | $0.13 | $0.43 / $0.87 | 54 | 1M | ||
59max | $1.60 | $1.25 / $4.25 | 128 | 1M |
GDPval-AA questions
What is GDPval-AA?›
Create files that satisfy real professional work requests. It has 220 tasks. Created by Artificial Analysis.
Which model is best on GDPval-AA?›
Claude Opus 5.5 leads GDPval-AA 2.1 with 68.3, ahead of Muse Wonder Full (67.9) and Claude Sonnet 5.5 (66.9).
What is the best open-weights model on GDPval-AA?›
DeepSeek V4.1 Flash 1017 with 62.3, then MiMo v2.6 Pro (59.2).
How is GDPval-AA scored?›
Crowd-BT Elo; pairwise comparisons fitted with crowd-bt (v1.1 briefcase / v2.1 gdpval); 1 run per task per task.