UTC

Llama-3.3-70B-Instruct

Meta · open weights · served by 50 providers · tool calling

meta/llama-3.3-70b-instruct

Weights
Open
Released
2024-12-06
Knowledge cutoff
2023-12
Context window
128K
Deprecation dates

When each platform deprecates Llama-3.3-70B-Instruct, as LiteLLM records it. Plan the move before the date: after it, the platform may stop serving the model or stop updating it.

  • Azure AI Foundrydeprecates 2027-01-31azure_ai/Llama-3.3-70B-Instruct
  • Amazon Bedrockdeprecates 2027-03-30meta.llama3-3-70b-instruct-v1:0successor: meta.llama2-70b-chat-v1
  • Cerebrasdeprecated 2026-02-16cerebras/llama-3.3-70bsuccessor: cerebras/gpt-oss-120b
  • Together AIdeprecated 2026-10-22together_ai/meta-llama/Llama-3.3-70B-Instruct-Turbosuccessor: together_ai/meta-models/Muse-Glimmer-30B
Every upcoming deprecation →
Hugging Face repository

meta-llama/Llama-3.3-70B-Instruct at commit 6f6073b423, last changed 2024-12-21

Licence
llama3.3
Access
Gated (approval)
Library
transformers
Formats
.pth .safetensors
  • 8 weight files use a pickle-based format, which can run code when loaded. Load the safetensors files instead.
  • No model signature file is published with the weights.
  • Built on meta-llama/Llama-3.1-70B.

What the Hub's scanners report

38 weight files, 38 with a scan result. No scanner reports an issue with them.

The model at this commit, its 38 weight files and its base models. The Hub shows file hashes of a gated model only after you accept its licence.

Facts from the Hugging Face Hub and its file scanners (Protect AI, Cisco, Hugging Face picklescan, VirusTotal, JFrog), as of 2026-11-18. ShieldMap does not copy model cards.

Who serves it
  • Abacusllama-3.3-70b-versatiledeprecated
  • Abacusmeta-llama/Meta-Llama-3.3-70B-Instructdeprecated
  • AI/ML APImeta-llama/Llama-3.3-70B-Instruct-Turbo
  • Amazon Bedrockmeta.llama3-3-70b-instruct-v1:0deprecated
  • Amazon Bedrockus.meta.llama3-3-70b-instruct-v1:0deprecated
  • CloudFerro Sherlockmeta-llama/Llama-3.3-70B-Instruct
  • Cloudflare Workers AI@cf/meta/llama-3.3-70b-instruct-fp8-fast
  • Cortecsllama-3.3-70b-instruct
  • Crusoemeta-llama/Llama-3.3-70B-Instructdeprecated
  • Deep Inframeta-llama/Llama-3.3-70B-Instruct-Turbo
  • Eden AIdeepinfra/meta-llama/Llama-3.3-70B-Instruct
  • Eden AIgreenpt/llama-3.3-70b-instruct
  • Eden AIionos/meta-llama/Llama-3.3-70B-Instruct
  • Eden AIscaleway/llama-3.3-70b-instruct
  • evrocnvidia/Llama-3.3-70B-Instruct-FP8
  • GreenPTllama-3.3-70b-instruct
  • Groqllama-3.3-70b-versatiledeprecated
  • Heliconellama-3.3-70b-instruct
  • Heliconellama-3.3-70b-versatile
  • Hugging Facemeta-llama/Llama-3.3-70B-Instruct
  • Inference APIsmeta-llama/Llama-3.3-70B-Instruct-Turbo
  • IO.NETmeta-llama/Llama-3.3-70B-Instruct
  • Kilo Gatewaymeta-llama/llama-3.3-70b-instruct
  • DevPass (LLM Gateway)llama-3.3-70b-instruct
26 more providers

LLM Gateway (novita/llama-3.3-70b-instruct), Meganova (meta-llama/Llama-3.3-70B-Instruct), Merge Gateway (meta/llama-3.3-70b-instruct), NanoGPT (TEE/llama3-3-70b), NanoGPT (meta-llama/llama-3.3-70b-instruct), Neon (meta-llama-3-3-70b-instruct), Novita AI (meta-llama/llama-3.3-70b-instruct), Obscurify.ai (meta-llama/llama-3.3-70b-instruct), OCI Generative AI (meta.llama-3.3-70b-instruct), OpenRouter (meta-llama/llama-3.3-70b-instruct), OVHcloud AI Endpoints (meta-llama-3_3-70b-instruct), Pendra (llama3.3:70b), Phala (meta-llama/llama-3.3-70b-instruct), Poe (cerebras/llama-3.3-70b-cs), Prime Inference (meta-llama/llama-3.3-70b-instruct), Regolo AI (llama-3.3-70b-instruct), Runcrate (meta-llama/Llama-3.3-70B-Instruct-Turbo), Scaleway (llama-3.3-70b-instruct), Snowflake Cortex (snowflake-llama3.3-70b), STACKIT (cortecs/Llama-3.3-70B-Instruct-FP8-Dynamic), Tinfoil (llama3-3-70b), Together AI (meta-llama/Llama-3.3-70B-Instruct-Turbo), Venice AI (llama-3.3-70b), Vercel AI Gateway (meta/llama-3.3-70b), CoreWeave (meta-llama/Llama-3.3-70B-Instruct), watsonx.ai (meta-llama/llama-3-3-70b-instruct)

Running it yourself

Open weights are loaded and served by code that has its own advisories. These inference servers and model loaders are on ShieldMap; load weights only from the publisher's own repository, and prefer safetensors over formats that can run code.

How supply-chain attacks on models work →

Sources: models.dev (MIT) and the LiteLLM model catalogue (MIT). Deprecation dates are as each platform published them and as LiteLLM records them; check the platform before you rely on one. Data as of 2026-11-22.