GraphCanon updated 2w · GitHub synced 2w
Decision brief
code-eval assesses large language models with the human-eval benchmark to provide insights into code generation reliability.
Good fit when
- When you need clear comparisons of pass rates for different LLMs using standardized tests
- If your project involves optimizing for edge cases in model tokenizing and loading processes
Avoid when
- If the tool's results do not correlate well with the official published benchmarks due to unknown prompt differences
- For real-time or dynamic evaluations as this repo offers pre-computed static results only
Observed Jul 17, 2026 · Source: enrich:decision_facts
Verify the decision
Maintenance and security
Full trust report- Maintenance
- Dormant (1058d since push)
- As of 2w
- Provenance
- Not a fork · Personal account
- As of 2w
- Security (OSV)
- 73 low (73 low)
- As of 1mo
Public GitHub metadata and optional OSV scans. Signals, not a guarantee. Trust methodology.
Install
pip install code-evalSimilar tools
Same-category neighbours. No typed graph edges are catalogued for this tool yet.
Evidence and technical details
Sourced facts, taxonomy, compatibility claims, README excerpt, and machine-readable endpoints.
Overview
A Python-based repository to evaluate the performance of large language models using the human-eval benchmark. It assesses different models and presents results based on their pass rates at various points, offering insights into model reliability in generating code solutions.
Capability facts
- Languages
- python
Source: github.language · Aug 5, 2026
Categories
Compatibility
Sourced claims from the README excerpt - not unsourced marketing copy.
Source: README excerpt (regex_v1, Aug 5, 2026)
Create python environmentSource link
Tags
README
code-eval
What
This is a repo I use to run human-eval on code models, adjust as needed. Some scripts were adjusted from wizardcoder repo (process_eval.py). The evaluation code is duplicated in several files, mostly to handle edge cases around model tokenizing and loading (will clean it up).
Results
Table is sorted by pass@1 score.
| model | size | pass@1 | pass@10 | screenshot |
|---|---|---|---|---|
| sahil2801/replit-code-instruct-glaive | 3B | 63.5% | 67% | |
| WizardCoder-15B-V1.0 | 15B | 57% | 68.9% | |
| bigcode/starcoder | 15B | 34.6% | 48.7% | |
| openchat/opencoderplus | 15B | 27.3% | 43.9% | |
| teknium/Replit-v1-CodeInstruct-3B | 3B | 25.8% | 42.6% | |
| teknium/Replit-v2-CodeInstruct-3B | 3B | 21.5% | 31% | |
| replit-code-v1-3b | 3B | 17.1% | 29.8% | |
| mpt-7b | 7B | 15.9% | 23.7% | |
| xgen-7b-8k-base | 7B | 14.9% | 22.5% | |
| openllama-7b-v2 | 7B | 14% | 23.1% | |
| llama-2-7b | 7B | 13.1% | 21.9% | |
| llama-7b | 7B | 12.1% | 18.9% | |
| mpt-30b | 30B | pending | pending | pending |
FAQ
Why is there a discrepancy on some of the scores between official numbers?
Because it is not obvious or published what prompt or processing the official models used to conduct their evaluation on this benchmark. The goal here is to try and best reproduce those numbers, in many cases it is possible to get very close to the published numbers.
All of the scores here were run independently of any published numbers and are reproducible by cloning the repo and following the setup.
Why do some models have a filter_code post generation step?
Base models can in many cases repeat outputs, breaking the benchmark scores. Instruct models don't have this problem and so you won't see this step, they tend to output a end of sequence token.
Setup
Create python environment
python -m venv env && source env/bin/activate
Install dependencies
pip install -r requirements.txt
For agents
This page has a .md twin and JSON over the API.