Qwen2.5-7B-Instruct
7.61B parameters · 28 layers · 4 KV heads. Compare weight precision and context budgets.
Open guide →MODEL DIRECTORY01 / MODEL MEMORY 02 / MODEL MEMORY 03 / MODEL MEMORY 04 / MODEL MEMORY 05 / MODEL MEMORY 06 / MODEL MEMORY
01 / MODEL → MEMORY 02 / MEMORY → MODELS 03 / CONTEXT → CACHE
Know the checkpoint you are sizing.
A focused catalog of six dense text models supported by our memory tools. Each entry links to the publisher, architecture configuration, and a dedicated planning guide.
Qwen2.5-14B-Instruct
14.7B parameters · 48 layers · 8 KV heads. Compare weight precision and context budgets.
Open guide →Qwen2.5-32B-Instruct
32.5B parameters · 64 layers · 8 KV heads. Compare weight precision and context budgets.
Open guide →Qwen3-8B
8.2B parameters · 36 layers · 8 KV heads. Compare weight precision and context budgets.
Open guide →Qwen3-14B
14.8B parameters · 40 layers · 8 KV heads. Compare weight precision and context budgets.
Open guide →DeepSeek-R1-Distill-Qwen-32B
32.5B parameters · 64 layers · 8 KV heads. Compare weight precision and context budgets.
Open guide →Specifications at a glance
| Checkpoint | Parameters | Layers | KV heads | Tool context limit | Source |
|---|---|---|---|---|---|
| Qwen2.5-7B-Instruct | 7.61B | 28 | 4 | 32,768 | Publisher ↗ |
| Qwen2.5-14B-Instruct | 14.7B | 48 | 8 | 32,768 | Publisher ↗ |
| Qwen2.5-32B-Instruct | 32.5B | 64 | 8 | 32,768 | Publisher ↗ |
| Qwen3-8B | 8.2B | 36 | 8 | 32,768 | Publisher ↗ |
| Qwen3-14B | 14.8B | 40 | 8 | 32,768 | Publisher ↗ |
| DeepSeek-R1-Distill-Qwen-32B | 32.5B | 64 | 8 | 131,072 | Publisher ↗ |
Go from specifications to a configuration
LLM VRAM calculator
Estimate weights, cache and runtime allocation for a named checkpoint, quantization and context.
Calculate memory →GPU model-fit finder
Enter your available VRAM and compare six model checkpoints with the same assumptions.
Find a memory fit →KV-cache calculator
See how context, parallel sequences and cache precision change conversation memory.
Size the cache →What belongs in this directory?
Each listed checkpoint has a linked source configuration and an explicit cache layout. This initial set covers dense grouped-query-attention text models. It is not a ranking, recommendation list, complete model catalog or claim that these are the newest releases.
DeepSeek-R1-Distill-Qwen-32B is listed separately from Qwen2.5-32B because the checkpoints, tokenizers and intended usage differ, even though their memory architecture inputs are similar. The full DeepSeek-R1 MoE is not supported by these tools.
Runtime references
Sources and review date
Publisher specifications reviewed October 31, 2026. Weight sizes and memory tables are InfoIsland calculations using the assumptions above; they are not measured benchmark results.
- Qwen2.5-7B-Instruct model card · architecture configuration
- Qwen2.5-14B-Instruct model card · architecture configuration
- Qwen2.5-32B-Instruct model card · architecture configuration
- Qwen3-8B model card · architecture configuration
- Qwen3-14B model card · architecture configuration
- DeepSeek-R1-Distill-Qwen-32B model card · architecture configuration
- llama.cpp server documentation: context, parallelism and cache types
- Ollama: context length and offloading