- Size
- 70.6B parameterscounted from the safetensors header on the Hub
- Licence
- unknown, Llama 3.3 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 3.3 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.
- On the Hub since
- 26 Nov 2024the date the repo was created
- Context length
- 128k tokensas an API provider reports it
- Fine-tune
- of meta-llama/Llama-3.1-70BLlamaForCausalLM
- On Hugging Face
- meta-llama/Llama-3.3-70B-Instruct344k downloads in the last 30 days, 3.2k likes
Scores
- default setting47.4%±2.927 Jan 2025, log
Epoch's name: Llama-3.3-70B-Instruct. Best closed model on this set: gpt-6.2-astra_max, 96.8% (OpenAI). All 341 results
- default setting41.6%±1.227 Jan 2025, log
Epoch's name: Llama-3.3-70B-Instruct. Best closed model on this set: gpt-5-2025-08-07_high, 98.1% (OpenAI). All 108 results
- default setting5.1%±2.425 Feb 2025, log
Epoch's name: Llama-3.3-70B-Instruct. Best closed model on this set: claude-fable-5_high, 100.0% (Anthropic). All 318 results
- default setting23.0%±1.79 Dec 2024, balrogai.com
Epoch's name: Llama-3.3-70B-Instruct. Best closed model on this set: gpt-6-astra_max, 68.3% (OpenAI). All 41 results
- default setting0.0%via Epoch AI
Epoch's name: Llama-3.3-70B-Instruct. Best closed model on this set: gemini-4-carbon_high, 36.0% (Google DeepMind). All 271 results
- default setting59.5%conceptualreasoning.ai
Epoch's name: Llama-3.3-70B-Instruct. Best closed model on this set: claude-fable-5-2_unknown, 99.2% (Anthropic). All 241 results
- default setting33.3%Fiction.live leaderboard
Epoch's name: Llama-3.3-70B-Instruct. Best closed model on this set: gpt-5-2025-08-07_medium, 97.2% (OpenAI). All 61 results
- default setting50.2%LiveBench Leaderboard
Epoch's name: Llama-3.3-70B-Instruct. Best closed model on this set: gemini-2.5-pro-exp-03-25, 82.3% (Google DeepMind). All 54 results
- default setting17.5%12.6 to 22.5conceptualreasoning.ai
Epoch's name: Llama-3.3-70B-Instruct. Best closed model on this set: claude-fable-5-2_xhigh, 70.6% (Anthropic). All 203 results
- Shots: 586.3%Phi-4 Technical Report
- Shots: 579.1%Stanford CRFM Leaderboard
Epoch's name: Llama-3.3-70B-Instruct. Best closed model on this set: gpt-4o-2024-11-20, 88.1% (OpenAI). All 217 results
- default setting19.9%SimpleBench Leaderboard
Epoch's name: Llama-3.3-70B-Instruct. Best closed model on this set: claude-fable-5-2_unknown, 94.9% (Anthropic). All 125 results
- default setting14.4%±0.0htihle.github.io
Epoch's name: Llama-3.3-70B-Instruct. Best closed model on this set: gpt-6-astra_promax, 93.6% (OpenAI). All 173 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: 2.5 GB, worked out from the model's config (kosbu/Llama-3.3-70B-Instruct-AWQ), 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
bartowski/Llama-3.3-70B-Instruct-GGUF (converted by bartowski), 219k downloads. Sizes are the real files on the Hub, all parts added up.
| Format | Size | Memory needed | Get it |
|---|---|---|---|
| IQ3_XXS | 25.6 GBfile size | 28.6 GB25.6 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:IQ3_XXSOllama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:IQ3_XXSDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-IQ3_XXS.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| IQ3_XS | 27.3 GBfile size | 30.3 GB27.3 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:IQ3_XSOllama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:IQ3_XSDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-IQ3_XS.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q3_K_S | 28.8 GBfile size | 31.8 GB28.8 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q3_K_SOllama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:Q3_K_SDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q3_K_S.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| IQ3_M | 29.7 GBfile size | 32.7 GB29.7 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:IQ3_MOllama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:IQ3_MDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-IQ3_M.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q3_K_M | 31.9 GBfile size | 34.9 GB31.9 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q3_K_MOllama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:Q3_K_MDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q3_K_M.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q3_K_L | 34.6 GBfile size | 37.6 GB34.6 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q3_K_LOllama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:Q3_K_LDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q3_K_L.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| IQ4_XS | 35.3 GBfile size | 38.3 GB35.3 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:IQ4_XSOllama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:IQ4_XSDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-IQ4_XS.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q3_K_XL | 35.4 GBfile size | 38.4 GB35.4 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q3_K_XLOllama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:Q3_K_XLDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q3_K_XL.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q4_0_4_4 | 37.2 GBfile size | 40.2 GB37.2 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q4_0_4_4Ollama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:Q4_0_4_4Download hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q4_0_4_4.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q4_0_4_8 | 37.2 GBfile size | 40.2 GB37.2 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q4_0_4_8Ollama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:Q4_0_4_8Download hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q4_0_4_8.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q4_0_8_8 | 37.2 GBfile size | 40.2 GB37.2 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q4_0_8_8Ollama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:Q4_0_8_8Download hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q4_0_8_8.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| IQ4_NL | 37.3 GBfile size | 40.3 GB37.3 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:IQ4_NLOllama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:IQ4_NLDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-IQ4_NL.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q4_0 | 37.4 GBfile size | 40.4 GB37.4 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q4_0Ollama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:Q4_0Download hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q4_0.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q4_K_S | 37.6 GBfile size | 40.6 GB37.6 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q4_K_SOllama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:Q4_K_SDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q4_K_S.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q4_K_M | 39.6 GBfile size | 42.6 GB39.6 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q4_K_MOllama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:Q4_K_MDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q4_K_M.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q4_K_L | 40.3 GBfile size | 43.3 GB40.3 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q4_K_LOllama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:Q4_K_LDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q4_K_L.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q5_K_S | 45.3 GBfile size | 48.3 GB45.3 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q5_K_SOllama ollama run hf.co/bartowski/Llama-3.3-70B-Instruct-GGUF:Q5_K_SDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q5_K_S.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q5_K_M | 46.5 GBfile size, 2 parts | 49.5 GB46.5 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q5_K_MDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q5_K_M/Llama-3.3-70B-Instruct-Q5_K_M-*-of-00002.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q5_K_L | 47.1 GBfile size, 2 parts | 50.1 GB47.1 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q5_K_LDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q5_K_L/Llama-3.3-70B-Instruct-Q5_K_L-*-of-00002.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q6_K | 53.9 GBfile size, 2 parts | 56.9 GB53.9 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q6_KDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q6_K/Llama-3.3-70B-Instruct-Q6_K-*-of-00002.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q6_K_L | 54.4 GBfile size, 2 parts | 57.4 GB54.4 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q6_K_LDownload hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q6_K_L/Llama-3.3-70B-Instruct-Q6_K_L-*-of-00002.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| Q8_0 | 69.8 GBfile size, 2 parts | 72.8 GB69.8 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:Q8_0Download hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-Q8_0/Llama-3.3-70B-Instruct-Q8_0-*-of-00002.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
| F16 | 131.4 GBfile size, 4 parts | 134.4 GB131.4 + 2.5 + 0.50 | Commandsllama.cpp llama-server -hf bartowski/Llama-3.3-70B-Instruct-GGUF:F16Download hf download bartowski/Llama-3.3-70B-Instruct-GGUF --include "Llama-3.3-70B-Instruct-f16/Llama-3.3-70B-Instruct-f16-*-of-00004.gguf" --local-dir Llama-3.3-70B-Instruct-GGUF |
MLX for Apple silicon
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| MLX 4bitmlx-community/Llama-3.3-70B-Instruct-4bit | 37.0 GBfile size | 40.0 GB37.0 + 2.5 + 0.50 | Download page |
| MLX 8bitmlx-community/Llama-3.3-70B-Instruct-8bit | 69.8 GBfile size | 72.8 GB69.8 + 2.5 + 0.50 | Download page |
Original weights size from the safetensors header, for vLLM, SGLang, Transformers
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| BF16original weights | 131.4 GBfile size | 134.4 GB131.4 + 2.5 + 0.50 |
hf download meta-llama/Llama-3.3-70B-Instruct --local-dir Llama-3.3-70B-InstructOther conversions (21; sizes not read)
- kosbu/Llama-3.3-70B-Instruct-AWQ AWQ, 86k downloads
- RedHatAI/Llama-3.3-70B-Instruct-FP8-dynamic FP8, 48k downloads
- unsloth/Llama-3.3-70B-Instruct-bnb-4bit bitsandbytes, 4bit, 8.4k downloads
- amd/Llama-3.3-70B-Instruct-FP8-KV FP8, 6.0k downloads
- ibnzterrell/Meta-Llama-3.3-70B-Instruct-AWQ-INT4 AWQ, int4, 4.1k downloads
- RedHatAI/Llama-3.3-70B-Instruct-quantized.w4a16 Other, w4a16, 2.3k downloads
- unsloth/Llama-3.3-70B-Instruct-unsloth-bnb-4bit bitsandbytes, 4bit, 1.1k downloads
- cortecs/Llama-3.3-70B-Instruct-FP8-Dynamic FP8, 1.1k downloads
- pearl-ai/Llama-3.3-70B-Instruct-pearl Other, 989 downloads
- RedHatAI/Llama-3.3-70B-Instruct-quantized.w8a8 Other, w8a8, 983 downloads
- amd/Llama-3.3-70B-Instruct-MXFP4-Preview Other, mxfp4, 974 downloads
- axana-labs/Llama-3.3-70B-Instruct-bnb-4bit bitsandbytes, 4bit, 272 downloads
- ionos/Meta-Llama-3.3-70B-Instruct-AWQ-INT4 AWQ, int4, 162 downloads
- unsloth/Llama-3.3-70B-Instruct-FP8-Dynamic FP8, 123 downloads
- Kaleto/Llama-3.3-70B-Instruct-NVFP4 NVFP4, 56 downloads
- cyankiwi/Llama-3.3-70B-Instruct-AWQ-INT4 AWQ, int4, 35 downloads
- Jacoby746/Llama-3.3-70B-Instruct-exl2-5.0bpw EXL2, 29 downloads
- Dracones/Llama-3.3-70B-Instruct_exl2_4.0bpw EXL2, 28 downloads
- Dracones/Llama-3.3-70B-Instruct_exl2_6.0bpw EXL2, 26 downloads
- Dracones/Llama-3.3-70B-Instruct_exl2_2.5bpw EXL2, 26 downloads
- MISHANM/meta-Llama-3.3-70B-Instruct-int8 bitsandbytes, int8, 23 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
8 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 |
|---|---|---|---|---|---|---|---|---|---|
| DeepInfra | $0.10 | $0.32 | 128k | fp8 | 19 | 0.5 s | 99.4% | yes | yes |
| novita | $0.14 | $0.40 | 12k | not given | 50 | 0.6 s | ? | yes | no |
| Parasail | $0.22 | $0.50 | 128k | fp8 | 58 | 0.6 s | 99.9% | no | yes |
| AkashML | $0.20 | $0.52 | 128k | fp8 | 19 | 1.4 s | 99.6% | yes | yes |
| CoreWeave | $0.71 | $0.71 | 125k | fp16 | 79 | 0.3 s | 100.0% | yes | yes |
| Groq | $0.59 | $0.79 | 128k | not given | 223 | 0.5 s | 100.0% | yes | no |
| Cloudflare | $0.29 | $2.25 | 23k | fp8 | 40 | 0.9 s | 99.4% | 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-3.3-70B-Instruct:deepinfra", "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-3.3-70B-Instruct70.6B37kUnreviewed
- nvidia/Llama-3.3-Nemotron-70B-Reward70.6B2.3kUnreviewed
- nvidia/Llama-3.3-Nemotron-70B-Reward-Principle70.6B2.2kUnreviewed
- nvidia/Llama-3.3-Nemotron-70B-Select70.6B241Unreviewed
- nvidia/Llama-3.3-Nemotron-70B-Feedback70.6B237Unreviewed
- nvidia/Llama-3.3-Nemotron-70B-Edit70.6B231Unreviewed
- nvidia/Llama-3.3-Nemotron-70B-Reward-Multilingual70.6B197Unreviewed
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.