# Coding Composite Index

> Multi-file software engineering, command-line recovery, repository tasks, and scientific coding.

- Canonical URL: https://prochoice.industries/coding
- Domain JSON endpoint: https://prochoice.industries/api/v1/domains/coding
- Epoch version: `coding-v1.0.0`
- Snapshot hash: `43237d0f0a7497fc`
- Updated: 2026-11-21
- Cost axis: Artificial Analysis Intelligence Index cost per task (USD, log scale)
- Evaluated entries: 137

## Headline picks

- Budget: GPT-6 Luna at 30.6 and $0.0051
- Best value: GPT-6.2 Sol at 64.2 and $0.41
- Top score: Claude Opus 5.5 at 77.5 and $10.0

## Pareto frontier

| Rank | Model | Effort | Provider | Score | Cost / task |
| :--- | :--- | :--- | :--- | :--- | :--- |
| 1 | Claude Opus 5.5 | Max | Anthropic | 77.5 | $10.0 |
| 2 | Claude Sonnet 5.5 | Max | Anthropic | 74.4 | $10.0 |
| 3 | Gemini 4 Carbon | High | Google | 71.7 | $7.76 |
| 4 | GPT-6.2 Astra | Extra high | OpenAI | 70.6 | $2.52 |
| 5 | GPT-6.1 Terra | Max | OpenAI | 68.8 | $0.96 |
| 6 | GPT-6.2 Sol | Extra high | OpenAI | 65.2 | $0.65 |
| 7 | GPT-6.2 Sol | Medium | OpenAI | 64.2 | $0.41 |
| 8 | GPT-6.1 Terra | Extra high | OpenAI | 58.2 | $0.40 |
| 9 | GPT-6.1 Terra | High | OpenAI | 57.6 | $0.26 |
| 10 | GPT-6.1 Terra | Medium | OpenAI | 54.9 | $0.17 |
| 11 | Muse Ember | Max | Meta | 52.2 | $0.14 |
| 12 | Muse Ember | Extra high | Meta | 47.8 | $0.12 |
| 13 | GPT-6.1 Terra | Low | OpenAI | 47.3 | $0.071 |
| 14 | GPT-6 Luna | Medium | OpenAI | 38.5 | $0.042 |
| 15 | GPT-6 Luna | Low | OpenAI | 30.6 | $0.0051 |

## Active benchmarks

- FrontierCode-AA 2.0 (2.0, ACTIVE): https://artificialanalysis.ai/evaluations/frontiercode-2-0
- Terminal-Bench 4.0 (4.0, ACTIVE): https://artificialanalysis.ai/leaderboards/models
- SciCode (1.0, ACTIVE): https://artificialanalysis.ai/evaluations/scicode
- LiveBench 2.0 Coding (20261106, ACTIVE): https://livebench.ai/
- Epoch MirrorCode (2026.11, ACTIVE): https://epoch.ai/benchmarks/mirrorcode
- Sakana ALE-Bench (2026.11, ACTIVE): https://github.com/SakanaAI/ALE-Bench
- Datacurve DeepSWE v1.1 (1.1, ACTIVE): https://deepswe.datacurve.ai/
- SWE-bench Pro Public v2 (2.0, ACTIVE): https://scale.com/leaderboard/swe_bench_pro_public_v2
- FrontierSWE (2026.11, ACTIVE): https://www.frontierswe.com/

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