# Research Composite Index

> Document analysis, long-context synthesis, scientific reasoning, mathematics, and knowledge work.

- Canonical URL: https://prochoice.industries/research
- Domain JSON endpoint: https://prochoice.industries/api/v1/domains/research
- Epoch version: `research-v1.0.0`
- Snapshot hash: `43237d0f0a7497fc`
- Updated: 2026-11-21
- Cost axis: Artificial Analysis Intelligence Index cost per task (USD, log scale)
- Evaluated entries: 210

## Headline picks

- Budget: GPT-6 Luna at 35.0 and $0.0051
- Best value: GPT-6.2 Sol at 71.3 and $0.56
- Top score: Claude Fable 5.2 at 81.7 and $26.5

## Pareto frontier

| Rank | Model | Effort | Provider | Score | Cost / task |
| :--- | :--- | :--- | :--- | :--- | :--- |
| 1 | Claude Fable 5.2 | Max | Anthropic | 81.7 | $26.5 |
| 2 | Claude Fable 5.2 | Extra high | Anthropic | 80.3 | $14.8 |
| 3 | Claude Fable 5.2 | High | Anthropic | 79.9 | $9.08 |
| 4 | GPT-6.2 Astra | Extra high | OpenAI | 77.8 | $2.52 |
| 5 | GPT-6.2 Astra | High | OpenAI | 76.7 | $2.07 |
| 6 | GPT-6.2 Astra | Medium | OpenAI | 73.3 | $1.62 |
| 7 | GPT-6.2 Sol | Extra high | OpenAI | 71.8 | $0.65 |
| 8 | GPT-6.2 Sol | High | OpenAI | 71.3 | $0.56 |
| 9 | GPT-6.2 Sol | Medium | OpenAI | 68.1 | $0.41 |
| 10 | GPT-6.1 Terra | Extra high | OpenAI | 59.6 | $0.40 |
| 11 | GPT-6.2 Sol | Low | OpenAI | 58.0 | $0.18 |
| 12 | GPT-6.1 Terra | Medium | OpenAI | 54.0 | $0.17 |
| 13 | GPT-6.1 Terra | Low | OpenAI | 48.7 | $0.071 |
| 14 | GPT-6 Luna | Medium | OpenAI | 39.3 | $0.042 |
| 15 | GPT-6 Luna | Low | OpenAI | 35.0 | $0.0051 |

## Active benchmarks

- LMArena Document Analysis (document_style_control, ACTIVE): https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset
- AA-Briefcase (1.1, ACTIVE): https://artificialanalysis.ai/evaluations/aa-briefcase
- GDPval-AA (2.1, ACTIVE): https://artificialanalysis.ai/evaluations/gdpval-aa
- AA Multi-Document Long Context (1.0, ACTIVE): https://artificialanalysis.ai/leaderboards/models
- AA Omniscience Accuracy (1.1, ACTIVE): https://artificialanalysis.ai/leaderboards/models
- GDP-PDF Synthesis (1.0, ACTIVE): https://artificialanalysis.ai/leaderboards/models
- HLE-Diamond (1.0, ACTIVE): https://artificialanalysis.ai/leaderboards/models
- CritPt v2 Hard (2.0, ACTIVE): https://artificialanalysis.ai/leaderboards/models
- Terminal-Bench-Science (0.1, ACTIVE): https://artificialanalysis.ai/evaluations/terminal-bench-science
- AA Analyst Agent (1.0, ACTIVE): https://artificialanalysis.ai/evaluations/aa-analyst-agent
- AA Medical Knowledge Assessment (1.0, ACTIVE): https://artificialanalysis.ai/leaderboards/models
- AA VIALS Scientific Lab (1.0, ACTIVE): https://artificialanalysis.ai/leaderboards/models
- Epoch SimpleQA Verified (2026.11, ACTIVE): https://epoch.ai/benchmarks/simpleqa-verified
- Epoch FrontierMath Tiers 1-3 (2026.11, ACTIVE): https://epoch.ai/frontiermath
- Epoch FrontierMath Tier 4 (2026.11, ACTIVE): https://epoch.ai/frontiermath/tier-4
- Vals ProofBench (2026.11, ACTIVE): https://www.vals.ai/benchmarks/proof_bench
- ARC-AGI-2 Public Eval (2026.11, ACTIVE): https://arcprize.org/leaderboard
- Mercor APEX-Agents (2026.11, ACTIVE): https://www.mercor.com/apex-agents
- LMArena Factuality (text_factuality, ACTIVE): https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset
- LiveBench 2.0 Research (20261106, ACTIVE): https://livebench.ai/
- LiveBench 2.0 Data Analysis (20261106, ACTIVE): https://livebench.ai/

Full JSON dataset: https://prochoice.industries/api/v1/snapshot
