featurebench 0.2.0

FeatureBench Pipeline - A test-driven data generation pipeline for building and evaluating feature-level coding benchmarks

---

FeatureBench is a test-driven data generation and evaluation pipeline for feature-level coding benchmarks. It provides a unified CLI to run inference, evaluation, and dataset generation.

📰 News

📊 2026.05.18: We added lite split evaluation results for frontier models including GPT-5.5, Claude Opus 4.7, DeepSeek-V4, GLM-5.1, Kimi-2.6, Mimo-V2.5-Pro, and more to the leaderboard.

🚀 2026.03.27: We released the fast split containing 100 instances (a subset of full split). These instances require no GPU and are optimized for rapid evaluation. On an Intel Xeon Platinum 8457C with 944GB RAM, the average evaluation time per instance using gold patches is 57.2 seconds.

🎁 2026.02.06: We now support one-click inference for mainstream agent frameworks, including OpenHands, Claude Code, Codex, Gemini CLI, and mini-swe-agent. All supported agent frameworks can be found here. We have also open-sourced the FeatureBench data pipeline.

🏆 Leaderboard

Full interactive leaderboard with tabs, filters, and sorting.

Lite split results, ranked by %PASSED

Rank	Model	Scaffold	%PASSED	%RESOLVED
1	Claude Opus 4.7	OpenHands	78.2	46.7
2	GPT-5.5	OpenHands	69.8	26.7
3	Claude Opus 4.6	OpenHands	69.5	20
4	Claude Opus 4.5	OpenHands	67.2	20
5	GPT-5.4	OpenHands	66.2	23.3
6	GPT-5.1-Codex	Codex	60.2	20
7	DeepSeek-V4-Pro	OpenHands	59.6	26.7
8	Claude Opus 4.5	Claude Code	59.1	20
9	Kimi-2.6	OpenHands	49.4	20
10	Mimo-V2.5-Pro	OpenHands	47.8	13.3
11	Gemini-3-Pro-Preview	OpenHands	45.1	10
12	GLM-5.1	OpenHands	44.2	13.3
13	Gemini-3-Pro-Preview	Gemini-CLI	43.4	10
14	DeepSeek-V4-Flash	OpenHands	41.9	16.7
15	MiniMax M2.1	Mini-SWE-Agent	41.9	10
16	GLM 4.7	Mini-SWE-Agent	41.2	6.7
17	Qwen3-Coder-480B-A35B-Instruct	OpenHands	38.3	6.7
18	DeepSeek V3.2	OpenHands	35.9	6.7
19	Qwen3-Coder-30B-A3B-Instruct	OpenHands	23	3.3

🚀 Quickstart

Prerequisites:

• uv for Python environment management
• docker for reproducible builds and evaluation

# pypi
pip install featurebench
# or uv add featurebench

# local
git clone https://github.com/LiberCoders/FeatureBench.git
cd FeatureBench
uv sync
source .venv/bin/activate

Configure:

cp config_example.toml config.toml

See docs/config.md for a comprehensive reference (harness, infer, data pipeline) with examples.

Optional: pre-pull images to reduce network variance:

fb pull --mode lite                 # lite split image list (13 images)
fb pull --mode fast                 # fast split image list (18 images)
fb pull --mode full                 # full split image list (24 images)
fb pull --mode /path/to/images.txt  # one image name per line

# full list: featurebench/resources/constants/full_images.txt
# lite list: featurebench/resources/constants/lite_images.txt
# fast list: featurebench/resources/constants/fast_images.txt

Run inference:

fb infer \
    --config-path config.toml \
    --agent mini_swe_agent \
    --model openai/qwen3-coder-480b-a35b-instruct \
    --split fast

Run evaluation:

fb eval \
    -p runs/<timestamp>/output.jsonl \
    --split fast
    # use -p gold to verify the gold patches

🧭 CLI Overview

fb provides three core commands:

• fb infer runs featurebench.infer.run_infer (docs: docs/infer_cli_arg.md)
• fb eval runs featurebench.harness.run_evaluation (docs: docs/harness_cli_arg.md)
• fb data runs featurebench.pipeline (docs: docs/pipeline.md)

✍️ Citation

If you found FeatureBench useful, please cite us as:

@article{zhou2026featurebench,
  title={FeatureBench: Benchmarking Agentic Coding for Complex Feature Development},
  author={Zhou, Qixing and Zhang, Jiacheng and Wang, Haiyang and Hao, Rui and Wang, Jiahe and Han, Minghao and Yang, Yuxue and Wu, Shuzhe and Pan, Feiyang and Fan, Lue and others},
  journal={arXiv preprint arXiv:2602.10975},
  year={2026}
}

📧 Contact

If you have any questions, feel free to contact [e-mail] or [e-mail].