reasoning benchmark
BigCodeBench
A benchmark that challenges LLMs to invoke multiple function calls as tools from 139 libraries and 7 domains for 1,140 fine-grained programming tasks. Evaluates code generation with diverse function calls and complex instructions, featuring two variants: Complete (code completion based on comprehensive docstrings) and Instruct (generating code from natural language instructions).
Updated Aug 7, 2026
Published models2
Registry coverage2
MetricScore
EvidenceB
BigCodeBench leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | GO | 45.4% | 100.0% | 2 | C | |
| 2 | AC | 41.0% | 0.0% | 2 | C |
Score distribution
Top published rows on the benchmark's original scale.
BigCodeBench highlights
The top published results on this benchmark's own scale.
What is BigCodeBench?
Definition and scoring fields from the benchmark registry.
A benchmark that challenges LLMs to invoke multiple function calls as tools from 139 libraries and 7 domains for 1,140 fine-grained programming tasks. Evaluates code generation with diverse function calls and complex instructions, featuring two variants: Complete (code completion based on comprehensive docstrings) and Instruct (generating code from natural language instructions).
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- BigCodeBench
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- bigcodebench|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about BigCodeBench.
Which model scores highest on BigCodeBench?
Gemini Diffusion is currently ranked first with 45.4%.
What does BigCodeBench measure?
A benchmark that challenges LLMs to invoke multiple function calls as tools from 139 libraries and 7 domains for 1,140 fine-grained programming tasks. Evaluates code generation with diverse function calls and complex instructions, featuring two variants: Complete (code completion based on comprehensive docstrings) and Instruct (generating code from natural language instructions).
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
2 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.