- Size
- 3.2B parameterscounted from the safetensors header on the Hub
- Licence
- unknown, Llama 3.2 Community License AgreementOur reading, check the licence: Commercial use is allowed unless your products had more than 700 million monthly active users in the calendar month before the Llama 3.2 release date, in which case you need a licence from Meta. If you distribute it, or a product or service containing it, you must include the agreement and display "Built with Llama", and any AI model built with Llama or its outputs must start its name with "Llama". The Acceptable Use Policy is part of the licence but is a separate document, not in this text, so any extra limits it sets are not covered here.
- On the Hub since
- 18 Sep 2024the date the repo was created
- Context length
- 128k tokensas an API provider reports it
- Kind
- Base model (no base model declared on the Hub)LlamaForCausalLM
- On Hugging Face
- meta-llama/Llama-3.2-3B-Instruct820k downloads in the last 30 days, 3.0k likes
Scores
- default setting10.1%±1.311 Nov 2024, balrogai.com
Epoch's name: Llama-3.2-3B-Instruct. Best closed model on this set: gpt-6-astra_max, 68.3% (OpenAI). All 41 results
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.
Independent scores from Epoch AI (CC-BY 4.0), fetched 19 Nov 2026: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. Each row links its own source. What the badges mean.
Run it yourself
Memory needed = weights + context cache + 0.50 GB for the runtime. Context cache at 8k tokens: 0.88 GB, worked out from the model's config (RedHatAI/Llama-3.2-3B-Instruct-FP8-dynamic), at 16-bit precision. Choose a machine above to see a verdict for each format. Method
GGUF for llama.cpp, Ollama, LM Studio and others
unsloth/Llama-3.2-3B-Instruct-GGUF (converted by unsloth), 206k downloads. Sizes are the real files on the Hub, all parts added up.
| Format | Size | Memory needed | Get it |
|---|---|---|---|
| UD-IQ3_XXS | 1.3 GBfile size | 2.7 GB1.3 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:UD-IQ3_XXSOllama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:UD-IQ3_XXSDownload hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-UD-IQ3_XXS.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| Q3_K_S | 1.4 GBfile size | 2.8 GB1.4 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:Q3_K_SOllama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:Q3_K_SDownload hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-Q3_K_S.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| Q3_K_M | 1.6 GBfile size | 2.9 GB1.6 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:Q3_K_MOllama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:Q3_K_MDownload hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-Q3_K_M.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| UD-Q3_K_XL | 1.6 GBfile size | 3.0 GB1.6 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:UD-Q3_K_XLOllama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:UD-Q3_K_XLDownload hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-UD-Q3_K_XL.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| IQ4_XS | 1.7 GBfile size | 3.1 GB1.7 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:IQ4_XSOllama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:IQ4_XSDownload hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-IQ4_XS.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| IQ4_NL | 1.8 GBfile size | 3.2 GB1.8 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:IQ4_NLOllama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:IQ4_NLDownload hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-IQ4_NL.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| Q4_0 | 1.8 GBfile size | 3.2 GB1.8 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:Q4_0Ollama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:Q4_0Download hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-Q4_0.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| Q4_K_S | 1.8 GBfile size | 3.2 GB1.8 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:Q4_K_SOllama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:Q4_K_SDownload hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-Q4_K_S.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| Q4_K_M | 1.9 GBfile size | 3.3 GB1.9 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:Q4_K_MOllama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:Q4_K_MDownload hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-Q4_K_M.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| UD-Q4_K_XL | 1.9 GBfile size | 3.3 GB1.9 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:UD-Q4_K_XLOllama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:UD-Q4_K_XLDownload hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-UD-Q4_K_XL.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| Q4_1 | 1.9 GBfile size | 3.3 GB1.9 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:Q4_1Ollama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:Q4_1Download hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-Q4_1.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| Q5_K_S | 2.1 GBfile size | 3.5 GB2.1 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:Q5_K_SOllama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:Q5_K_SDownload hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-Q5_K_S.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| Q5_K_M | 2.2 GBfile size | 3.5 GB2.2 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:Q5_K_MOllama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:Q5_K_MDownload hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-Q5_K_M.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| UD-Q5_K_XL | 2.2 GBfile size | 3.5 GB2.2 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:UD-Q5_K_XLOllama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:UD-Q5_K_XLDownload hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-UD-Q5_K_XL.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| Q6_K | 2.5 GBfile size | 3.8 GB2.5 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:Q6_KOllama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:Q6_KDownload hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-Q6_K.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| UD-Q6_K_XL | 2.8 GBfile size | 4.1 GB2.8 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:UD-Q6_K_XLOllama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:UD-Q6_K_XLDownload hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-UD-Q6_K_XL.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| Q8_0 | 3.2 GBfile size | 4.6 GB3.2 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:Q8_0Ollama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:Q8_0Download hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-Q8_0.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| UD-Q8_K_XL | 3.9 GBfile size | 5.3 GB3.9 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:UD-Q8_K_XLOllama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:UD-Q8_K_XLDownload hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-UD-Q8_K_XL.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| F16 | 6.0 GBfile size | 7.4 GB6.0 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:F16Ollama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:F16Download hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-F16.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
| BF16 | 6.0 GBfile size | 7.4 GB6.0 + 0.88 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Llama-3.2-3B-Instruct-GGUF:BF16Ollama ollama run hf.co/unsloth/Llama-3.2-3B-Instruct-GGUF:BF16Download hf download unsloth/Llama-3.2-3B-Instruct-GGUF --include "Llama-3.2-3B-Instruct-BF16.gguf" --local-dir Llama-3.2-3B-Instruct-GGUF |
Original weights size from the safetensors header, for vLLM, SGLang, Transformers
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| BF16original weights | 6.0 GBfile size | 7.4 GB6.0 + 0.88 + 0.50 |
hf download meta-llama/Llama-3.2-3B-Instruct --local-dir Llama-3.2-3B-InstructOther conversions (13; sizes not read)
- RedHatAI/Llama-3.2-3B-Instruct-FP8-dynamic FP8, 51k downloads
- unsloth/Llama-3.2-3B-Instruct-unsloth-bnb-4bit bitsandbytes, 4bit, 32k downloads
- mobilint/Llama-3.2-3B-Instruct Other, 28k downloads
- mlc-ai/Llama-3.2-3B-Instruct-q4f32_1-MLC Other, 27k downloads
- unsloth/Llama-3.2-3B-Instruct-bnb-4bit bitsandbytes, 4bit, 26k downloads
- mlc-ai/Llama-3.2-3B-Instruct-q4f16_1-MLC Other, 8.1k downloads
- mlc-ai/Llama-3.2-3B-Instruct-q4f16_0-MLC Other, 2.2k downloads
- RedHatAI/Llama-3.2-3B-Instruct-FP8 FP8, 1.8k downloads
- fbaldassarri/meta-llama_Llama-3.2-3B-Instruct-auto_round-int4-gs128-asym GPTQ, int4, 366 downloads
- unsloth/Llama-3.2-3B-Instruct-FP8-Dynamic FP8, 198 downloads
- simaai/Llama-3.2-3B-Instruct-Autoround-Safetensors GPTQ, 170 downloads
- unsloth/Llama-3.2-3B-Instruct-FP8-Block FP8, 157 downloads
- bnb-community/Llama-3.2-3B-Instruct-bnb-4bit bitsandbytes, 4bit, 131 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
3 providers serve this model through the Hugging Face router. Prices are US dollars per million tokens. Precision is what the provider reports; a lower precision than the original weights can change answers. Live first, then cheapest output.
| Provider | In | Out | Context | Precision | Tokens/s | First token | Uptime 24h | Tools | JSON |
|---|---|---|---|---|---|---|---|---|---|
| Parasail | $0.05 | $0.33 | 128k | bf16 | 68 | 0.3 s | 100.0% | no | yes |
| Cloudflare | $0.051 | $0.34 | 78k | not given | 73 | 0.3 s | 99.8% | no | no |
| featherless-ai | ? | ? | ? | not given | ? | ? | ? | not given | not given |
Through the router, with a Hugging Face token (the router bills your Hugging Face account). Change the name after the colon to pick another provider.
curl https://router.huggingface.co/v1/chat/completions \
-H "Authorization: Bearer $HF_TOKEN" -H "Content-Type: application/json" \
-d '{"model": "meta-llama/Llama-3.2-3B-Instruct:parasail", "messages": [{"role": "user", "content": "Hello"}]}'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
Repos that declare this model as their base, most downloaded first. Listing is not a recommendation; unreviewed ones have not been checked.
- unsloth/Llama-3.2-3B-Instruct3.2B480kUnreviewed
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.