|
Download README.md from These-Guys-Know/tgk-ai-coding-harnesses: direct link, hf CLI and curl.
- Browser
- Download file 2.45 kB
-
https://huggingface.co/datasets/These-Guys-Know/tgk-ai-coding-harnesses/resolve/main/README.md
- Command line
-
hf download hf://datasets/These-Guys-Know/tgk-ai-coding-harnesses/README.md
-
curl -L -o README.md https://huggingface.co/datasets/These-Guys-Know/tgk-ai-coding-harnesses/resolve/main/README.md
2.45 kB
| pretty_name: TGK AI Coding Harnesses | |
| # TGK AI Coding Harnesses | |
| Article: https://theseguysknow.io/best-ai-coding-harnesses/ | |
| Version: v2 (18 November 2026). Zenodo DOI: https://doi.org/10.5281/zenodo.24348696. Version 1: https://doi.org/10.5281/zenodo.24316605. | |
| 18 tier-list entries, including the numbered Top 10, unchanged since 5 October 2026. Licence facts for the ranked ten are taken from the TGK snapshot; later checks for the other eight are dated 17 November 2026. | |
| New in v2: `third_party_benchmarks.csv`, the 78 benchmark figures the article cites from Artificial Analysis Coding Agent Index v2.1, Runta FrontierHarness v1.1 and Joel Niklaus's Harness Leaderboard v0.1.8, all checked on 18 November 2026. The FrontierHarness per-model pass rates were computed by TGK from Runta's published trial records and are marked as derived. | |
| ## Limitations | |
| This is an editorial preference dataset, not a controlled coding benchmark. The tiers come from TGK group experience. Benchmark figures and licence data are third-party information linked to their sources and are not TGK research. Benchmark scores from different sources use different tasks, models and settings, so they should not be averaged or compared across sources. | |
| ## Results | |
| | test_date | rank | model | tier | | |
| | --- | --- | --- | --- | | |
| | 2026-10-05 | 1 | DeepSeek Harness | S+ | | |
| | 2026-10-05 | 2 | Hermes | S | | |
| | 2026-10-05 | 3 | Factory Droid | S | | |
| | 2026-10-05 | 4 | Devin | S | | |
| | 2026-10-05 | | CodeWhale | S | | |
| | 2026-10-05 | 5 | Oh My Pi | A | | |
| | 2026-10-05 | 6 | Pi Agent | A | | |
| | 2026-10-05 | 7 | Prime Agent | A | | |
| | 2026-10-05 | 8 | Grok Build | B | | |
| | 2026-10-05 | 9 | Codex | B | | |
| | 2026-10-05 | 10 | Cursor Agent | B | | |
| | 2026-10-05 | | Claude Code | B | | |
| | 2026-10-05 | | Command Code | C | | |
| | 2026-10-05 | | OpenCode | C | | |
| | 2026-10-05 | | ZCode | D | | |
| | 2026-10-05 | | Kimi Code | D | | |
| | 2026-10-05 | | Qwen Code | E | | |
| | 2026-10-05 | | Antigravity | E | | |
| ## Files | |
| - `model_results.csv`, `evidence_records.csv`: tiers, ranks and licence checks (unchanged from v1). | |
| - `third_party_benchmarks.csv`: cited benchmark figures (new in v2). | |
| - `test_prompts.csv`: empty; no prompts were used for the editorial ranking. | |
| ## Dataset mirrors and citation | |
| Hugging Face: https://huggingface.co/datasets/These-Guys-Know/tgk-ai-coding-harnesses | |
| Kaggle: https://www.kaggle.com/datasets/mikehazard/tgk-ai-coding-harnesses | |
| Zenodo: https://doi.org/10.5281/zenodo.24348696 | |