Qwen2.5 Coder 7B Instruct
Alibaba Qwen
The practical local copilot. Supports fill-in-the-middle, so it works as an inline completion model rather than only a chat assistant.
Good fit
- Weights
- 14 GB
- KV cache
- 440 MB
- Overhead
- 590 MB
- 15 GB of 23 GB usable VRAM.
- Chosen as the best quality that still fits a 32,768-token context (17 GB at that length).
- Fill-in-the-middle support
- Fast enough for inline completion
- Runs on 8 GB at Q4_K_M
- Weaker than general models outside code
- Repository-scale context needs the 32B
qwen2.5-coder:7bRunning Qwen2.5 Coder 7B Instruct on GeForce RTX 3090
| Quantization | Quality | Weights | KV cache | Total | ~tok/s | Fit |
|---|---|---|---|---|---|---|
| F16Pick | lossless | 14 GB | 440 MB | 15 GB | 48 | Good fit |
| Q8_0 | near-lossless | 7.5 GB | 440 MB | 8.6 GB | 90 | Excellent fit |
| Q6_K | near-lossless | 5.8 GB | 440 MB | 6.8 GB | 116 | Excellent fit |
| Q5_K_M | high | 5.0 GB | 440 MB | 6.0 GB | 134 | Excellent fit |
| Q4_K_M | balanced | 4.3 GB | 440 MB | 5.3 GB | 158 | Excellent fit |
| Q3_K_M | degraded | 3.5 GB | 440 MB | 4.5 GB | 195 | Excellent fit |
KV cache is sized at 8,192 tokens. Longer contexts cost proportionally more — the recommendation above reserves room for a working context.
You can fine-tune this here
- Base weights
- 4.0 GB
- Optimizer
- 602 MB
- Activations
- 890 MB
- Peak
- 7.0 GB
- 7.0 GB peak against 23 GB usable — room to raise batch size or sequence length.
- Base weights
- 14 GB
- Optimizer
- 602 MB
- Activations
- 890 MB
- Peak
- 17 GB
- 17 GB peak against 23 GB usable.
Where this model runs
VRAM 32 GB · F16 · 15 GB
VRAM 24 GB · F16 · 15 GB
VRAM 16 GB · Q8_0 · 8.6 GB
VRAM 16 GB · Q8_0 · 8.6 GB
VRAM 24 GB · F16 · 15 GB
VRAM 12 GB · Q6_K · 6.8 GB
VRAM 12 GB · Q6_K · 6.8 GB
VRAM 16 GB · Q8_0 · 8.6 GB
VRAM 48 GB · F16 · 15 GB
VRAM 80 GB · F16 · 15 GB
Unified 128 GB · F16 · 15 GB
Unified 48 GB · F16 · 15 GB
Unified 24 GB · Q8_0 · 8.6 GB
Unified 192 GB · F16 · 15 GB
Unified 16 GB · Q6_K · 6.8 GB
VRAM 24 GB · F16 · 15 GB
VRAM 16 GB · Q8_0 · 8.6 GB
VRAM 0 MB
VRAM 0 MB
88.4%
Qwen2.5-Coder model card
Reported by the model's author. ModelLM has not run these benchmarks and does not treat them as verified.
Qwen2.5 7B Instruct
7.6B · Apache 2.0
The default starting point for local work on 8–12 GB cards. Strong instruction following and reliable tool-call formatting for its size.
Qwen2.5 14B Instruct
14.8B · Apache 2.0
The sweet spot for 24 GB cards. Meaningfully stronger reasoning than 7B while still fine-tunable locally with QLoRA.
Qwen2.5 32B Instruct
32.8B · Apache 2.0
Approaches 70B quality at half the memory. Runs on a single 24 GB card at Q4_K_M with a modest context window.
Qwen2.5 72B Instruct
72.7B · Qwen License
Frontier-adjacent open weights. Needs a workstation, a multi-GPU rig or a large unified-memory Mac.
Catalogue figures come from each model’s published card. ModelLM has not independently measured them.