Task Shortlist · Research & analysis
Best API models for research & analysis
Go deep, and check the sources.
Trade-offs & Editorial Note
Fable leads the overall index, not every research task. Tool access and source checking still decide quality. Carbon is positioned for long, deep work even though Neon scores higher overall.
1
Claude Fable 5.2Anthropic
Hard reasoning and long-horizon agents.
Benchmark $26.54/task · $8.00 in / $40.00 out per 1M
64.6INDEX
Dossier 2
Difficult software, computer-use and document workflows.
Benchmark $4.00/task · $8.00 in / $40.00 out per 1M
62.1INDEX
Dossier 3
Gemini 4 CarbonGoogle
Deep engineering, mathematical and scientific reasoning.
Benchmark $7.76/task · $2.00 in / $10.00 out per 1M
60.0INDEX
Dossier 4
Complex coding, analytical reasoning and mixed-media workflows.
Benchmark $6.45/task · $3.00 in / $15.00 out per 1M
61.4INDEX
Dossier