Qwen · fine-tune of Qwen/Qwen2.5-3B
Qwen2.5-3B-Instruct
Runs on your machine
Choose a machine
We then show the largest format that fits, from real file sizes where we have them.
Scores
No independent scores for this model in Epoch AI's data. A score is only shown here when Epoch's name for a model matches this repo with certainty; an uncertain match is left out rather than guessed.
Claimed by the maker Claimed
The model card has no claimed scores in a machine-readable form (the card's model-index), so none are shown. The maker may give scores in the card text: read the card.
Other views, which we link to and do not copy: look this model up on Artificial Analysis and the Arena leaderboard.
Run it yourself
Memory needed = weights + context cache + 0.50 GB for the runtime. Context cache at 8k tokens: 0.28 GB, worked out from the model's config (Qwen/Qwen2.5-3B-Instruct), at 16-bit precision. Choose a machine above to see a verdict for each format. Method
weights context cache runtime
GGUF for llama.cpp, Ollama, LM Studio and others
Qwen/Qwen2.5-3B-Instruct-GGUF (the maker's own conversion), 324k downloads. Sizes are the real files on the Hub, all parts added up. Showing the common levels.
| Format | Size | Memory needed | Memory | Get it |
|---|---|---|---|---|
| Q4_K_M | 2.0 GBfile size | 2.7 GB2.0 + 0.28 + 0.50 | Commandsllama.cpp llama-server -hf Qwen/Qwen2.5-3B-Instruct-GGUF:Q4_K_MOllama ollama run hf.co/Qwen/Qwen2.5-3B-Instruct-GGUF:Q4_K_MDownload hf download Qwen/Qwen2.5-3B-Instruct-GGUF --include "qwen2.5-3b-instruct-q4_k_m.gguf" --local-dir Qwen2.5-3B-Instruct-GGUF | |
| Q5_K_M | 2.3 GBfile size | 3.1 GB2.3 + 0.28 + 0.50 | Commandsllama.cpp llama-server -hf Qwen/Qwen2.5-3B-Instruct-GGUF:Q5_K_MOllama ollama run hf.co/Qwen/Qwen2.5-3B-Instruct-GGUF:Q5_K_MDownload hf download Qwen/Qwen2.5-3B-Instruct-GGUF --include "qwen2.5-3b-instruct-q5_k_m.gguf" --local-dir Qwen2.5-3B-Instruct-GGUF | |
| Q6_K | 2.6 GBfile size | 3.4 GB2.6 + 0.28 + 0.50 | Commandsllama.cpp llama-server -hf Qwen/Qwen2.5-3B-Instruct-GGUF:Q6_KOllama ollama run hf.co/Qwen/Qwen2.5-3B-Instruct-GGUF:Q6_KDownload hf download Qwen/Qwen2.5-3B-Instruct-GGUF --include "qwen2.5-3b-instruct-q6_k.gguf" --local-dir Qwen2.5-3B-Instruct-GGUF | |
| Q8_0 | 3.4 GBfile size | 4.1 GB3.4 + 0.28 + 0.50 | Commandsllama.cpp llama-server -hf Qwen/Qwen2.5-3B-Instruct-GGUF:Q8_0Ollama ollama run hf.co/Qwen/Qwen2.5-3B-Instruct-GGUF:Q8_0Download hf download Qwen/Qwen2.5-3B-Instruct-GGUF --include "qwen2.5-3b-instruct-q8_0.gguf" --local-dir Qwen2.5-3B-Instruct-GGUF |
MLX for Apple silicon
| Format | Size | Memory needed | Memory | Repo |
|---|---|---|---|---|
| MLX 4bitlmstudio-community/Qwen2.5-3B-Instruct-MLX-4bit | 1.6 GBfile size | 2.4 GB1.6 + 0.28 + 0.50 | Download page | |
| MLX 8bitlmstudio-community/Qwen2.5-3B-Instruct-MLX-8bit | 3.1 GBfile size | 3.8 GB3.1 + 0.28 + 0.50 | Download page |
Original weights size from the safetensors header, for vLLM, SGLang, Transformers
| Format | Size | Memory needed | Memory | Repo |
|---|---|---|---|---|
| BF16original weights | 5.7 GBfile size | 6.5 GB5.7 + 0.28 + 0.50 |
hf download Qwen/Qwen2.5-3B-Instruct --local-dir Qwen2.5-3B-InstructOther conversions (9; sizes not read)
- Qwen/Qwen2.5-3B-Instruct-AWQ AWQ, by the maker, 1.1M downloads
- Qwen/Qwen2.5-3B-Instruct-GPTQ-Int4 GPTQ, int4, by the maker, 6.1k downloads
- Qwen/Qwen2.5-3B-Instruct-GPTQ-Int8 GPTQ, int8, by the maker, 826 downloads
- unsloth/Qwen2.5-3B-Instruct-unsloth-bnb-4bit bitsandbytes, 4bit, 72k downloads
- unsloth/Qwen2.5-3B-Instruct-bnb-4bit bitsandbytes, 4bit, 37k downloads
- mobilint/Qwen2.5-3B-Instruct Other, 27k downloads
- mlc-ai/Qwen2.5-3B-Instruct-q4f16_1-MLC Other, 4.8k downloads
- mlc-ai/Qwen2.5-3B-Instruct-q4f32_1-MLC Other, 688 downloads
- JongYeop/Qwen2.5-3B-Instruct-FP4-W4A4 NVFP4, 284 downloads
The Hub tags 100 or more quantised repos as conversions of this model; we read the most downloaded and show the maker's own first.
Run it by API
No provider on Hugging Face Inference Providers serves this model (as of 19 Nov 2026, 13:58 UTC). Other hosts may; we only list what the router reports.
Not built yet: one key for every open model, with BlockRun, and our own tested-quality score per provider. Until then, precision and measured speed are the best guide to a provider here.
Fine-tunes of this model
- nvidia/LocateAnything-3B3.8B32kUnreviewed
- unsloth/Qwen2.5-3B-Instruct3.1B16kUnreviewed
- 5CD-AI/Vintern-3B-beta3.7B8.3kUnreviewed
Planned for this page Not built yet
- Try it in private chat
- A link to Solana AI's confidential chat for the models it serves. This model is not listed as served there.
- Fine-tune it
- Fine-tuning a few base models from this page and publishing the result here.
- Creators and earnings
- Pages for people who publish fine-tunes and adapters of this model, and what they earn.
- Fund and bounties
- Bounties for new open models, paid from a public fund. How the fund will work.
- Contribute compute
- How to give compute to open model work, with Akash.
Model data from the Hugging Face Hub 19 Nov 2026, 13:43 UTC. The repo. Found a mistake on this page? Tell us.