Updated 2026-08-06 · ranked by real download data
Best open coding models
Ranked nightly from our download snapshots of the Hugging Face catalog. Every entry shows its licence and what hardware it realistically needs.
| # | Model | Params | Context | Commercial use | 30d | Min VRAM |
|---|---|---|---|---|---|---|
| 01 | Qwen3-Coder-30B-A3B-Instruct-GGUF | — | — | ✓ apache-2.0 | 4.7 M | from 12.9 GB |
| 02 | Qwen2.5-Coder-14B-Instruct | 14.8 B | 33 K | ✓ apache-2.0 | 3.0 M | from 35.2 GB |
| 03 | Qwen3-Coder-Next-FP8 | 79.7 B | 262 K | ✓ apache-2.0 | 2.2 M | from 100.9 GB |
| 04 | Qwen2.5-Coder-7B-Instruct | 7.6 B | 33 K | ✓ apache-2.0 | 2.0 M | from 18.4 GB |
| 05 | Qwen2.5-Coder-32B-Instruct-AWQ | 32.8 B | 33 K | ✓ apache-2.0 | 1.7 M | from 26.7 GB |
| 06 | Qwen3-Coder-30B-A3B-Instruct | 30.5 B | 262 K | ✓ apache-2.0 | 1.5 M | from 72.3 GB |
| 07 | Qwen3-Coder-30B-A3B-Instruct-FP8 | 30.5 B | 262 K | ✓ apache-2.0 | 1.4 M | from 39.4 GB |
| 08 | Qwen2.5-Coder-32B-Instruct | 32.8 B | 33 K | ✓ apache-2.0 | 1.2 M | from 77.5 GB |
| 09 | Qwen2.5-Coder-7B | 7.6 B | 33 K | ✓ apache-2.0 | 1.0 M | from 18.4 GB |
| 10 | Qwen3-Coder-Next-FP8-dynamic | 79.8 B | 262 K | ✓ apache-2.0 | 864 K | from 102.4 GB |
| 11 | Qwen2.5-Coder-1.5B-Instruct | 1.5 B | 33 K | ✓ apache-2.0 | 728 K | from 4.1 GB |
| 12 | Qwen2.5-Coder-7B-Instruct-AWQ | 7.6 B | 33 K | ✓ apache-2.0 | 604 K | from 7.9 GB |
| 13 | deepseek-coder-7b-instruct-v1.5 | 6.9 B | 4 K | other | 599 K | from 16.7 GB |
| 14 | Qwen3-Coder-30B-A3B-Instruct-AWQ-4bit | 5.3 B | 262 K | ✓ apache-2.0 | 583 K | from 21.2 GB |
| 15 | DeepSeek-Coder-V2-Lite-Instruct | 15.7 B | 164 K | other | 557 K | from 37.4 GB |
| 16 | Qwen3-Coder-Next | 79.7 B | 262 K | ✓ apache-2.0 | 554 K | from 187.7 GB |
| 17 | deepseek-coder-6.7b-instruct | 6.7 B | 16 K | other | 553 K | from 16.3 GB |
| 18 | Qwen2.5-Coder-7B-Instruct-GPTQ-Int4 | 7.6 B | 33 K | ✓ apache-2.0 | 536 K | from 7.8 GB |
| 19 | Qwen2.5-Coder-7B-Instruct-AWQ | 7.6 B | 33 K | ✓ apache-2.0 | 522 K | from 7.8 GB |
| 20 | Qwen2.5-Coder-14B-Instruct-AWQ | 14.8 B | 33 K | ✓ apache-2.0 | 406 K | from 13.7 GB |
| 21 | CodeLlama-7b-hf | 6.7 B | 16 K | ⚠ llama2 | 356 K | from 16.3 GB |
| 22 | Qwen3-Coder-480B-A35B-Instruct-FP8 | 480.2 B | 262 K | ✓ apache-2.0 | 327 K | from 602.9 GB |
| 23 | gemma-4-12B-coder-fable5-composer2.5-v1-GGUF | — | — | ✓ apache-2.0 | 298 K | from 5.8 GB |
| 24 | Qwen3-Coder-30B-A3B-Instruct-AWQ | 30.5 B | 262 K | ✓ apache-2.0 | 254 K | from 23.6 GB |
| 25 | tiny_starcoder_py | 160 M | — | bigcode-openrail-m | 249 K | from 1.2 GB |
Membership and ranking refresh nightly after the snapshot run. VRAM is an estimate for the smallest quantization at 8K context — methodology.