Qwen2.5 Coder 7B Instruct

Qwen · released 2024-09-17 · apache-2.0 license

Qwen2.5-Coder-7B-Instruct is a code-specialized 7B model (Apache-2.0) with vendor LiveCodeBench 48.7, strong on pure code generation but weaker on general reasoning.

Key specs

TypeLocal open-weight
Parameters7.62B total
Architectureqwen2
Context window33K tokens
Knowledge cutoff
Modalitiestext
Recommended backends
Minimum viable rig

Benchmark scores

GPQA Diamond
SWE-bench Verified
AIME
MMLU-Pro
BFCL v3 (tool use)
Composite score
Community ratingNo reviews yet

VRAM & disk per quantization

QuantVRAMDiskRAMContext
Q4_K_M5.9 GB4.4 GB33K

Strengths & weaknesses

Strengths: Strong code generation for 7B (HumanEval ~88); Small enough for local/on-device use; Apache-2.0

Weaknesses: Weaker general reasoning than same-size general models; Limited agentic/SWE-bench results