reasoning benchmark
BigCodeBench-Full
A comprehensive benchmark that evaluates large language models' ability to solve complex, practical programming tasks via code generation. Contains 1,140 fine-grained tasks across 7 domains using function calls from 139 libraries. Challenges LLMs to invoke multiple function calls as tools and handle complex instructions for realistic software engineering and general-purpose reasoning tasks.
Updated Aug 7, 2026
Published models1
Registry coverage1
MetricScore
EvidenceB
BigCodeBench-Full leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | AC | 49.6% | 100.0% | 1 | C |
BigCodeBench-Full highlights
The top published results on this benchmark's own scale.
What is BigCodeBench-Full?
Definition and scoring fields from the benchmark registry.
A comprehensive benchmark that evaluates large language models' ability to solve complex, practical programming tasks via code generation. Contains 1,140 fine-grained tasks across 7 domains using function calls from 139 libraries. Challenges LLMs to invoke multiple function calls as tools and handle complex instructions for realistic software engineering and general-purpose reasoning tasks.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- BigCodeBench-Full
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- bigcodebench-full|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about BigCodeBench-Full.
Which model scores highest on BigCodeBench-Full?
Qwen2.5-Coder 32B Instruct is currently ranked first with 49.6%.
What does BigCodeBench-Full measure?
A comprehensive benchmark that evaluates large language models' ability to solve complex, practical programming tasks via code generation. Contains 1,140 fine-grained tasks across 7 domains using function calls from 139 libraries. Challenges LLMs to invoke multiple function calls as tools and handle complex instructions for realistic software engineering and general-purpose reasoning tasks.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
1 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.