Llama 3.3 70B
meta/llama-3.3-70b-instruct- Meta
- Family · Llama 3
- Open weights
- Text → Text
- Tools · JSON
Not on OSMEXO yet
Spec & list price
Context128K
Max output128K
Input / 1M$0.20
Output / 1M$0.522.6× input
Usage
Tokens / day4.1B▼ 25% 7d
Usage rank#134
Requests / day2.0M
Tool calls / req0.02
Per request · 30 days
Avg input1.5K0.1× median
Avg output1130.2× median
Cache hit25%0.3× median
Tool-call failures9.6%5.1× median
Consumption
Peak 9.8B 27 SepLow 3.2B 2 NovRank today #134
News & trends
1 mentions this week ▼ 80%Llama 3 family · videos, posts, articles · 30 days
News · latest from Meta
- Meta Networking at the 2026 OCP Global Summit: Open Innovation for Scale-Up, Scale-Out, and Scale-Across AI NetworksMeta Engineering · 26 Oct
- Scale Agentic Coding Workflows with Meta’s Muse Spark Open on NVIDIA GB300 NVL72NVIDIA · 15 Oct
- Meta Muse Spark Open is live on Together AITogether AI · 15 Oct
- Welcome Muse Spark Open: open-weight frontier multimodal mixture-of-experts model from Meta15 Oct
Benchmarks
Benchmark data: Artificial AnalysisOutput speed#244/33142 tok/s1.5× slower than median
Fastest provider216 tok/sbest of 8 providers
First token#900.66 s1.7× quicker than median
Family progress
Intelligence Index
- Family releases7.7
Technical details
- Tokenizer
- Llama3
- Knowledge cutoff
- 2023-12-31
- License
- llama3.3
- Parameters
- 70.6B
- Moderation
- none
- Downloads · likes
- 365.5K · 3.2K
Request parameters 17
- tools
- tool_choice
- response_format
- structured_outputs
- temperature
- top_p
- top_k
- min_p
- max_tokens
- stop
+7 more
- seed
- frequency_penalty
- logit_bias
- logprobs
- presence_penalty
- repetition_penalty
- top_logprobs
When to use
Best for
- General-purpose chat and writing
- Self-hosted deployments (open weights)
- Instruction-following assistants with tool calling
- Fine-tuning a capable open model
Think twice for
- Image input (text only)
- Very large outputs (short output cap)
- Very long documents (limited context)
Research papers
- Heads of Security: Learned Attention Gates Improve Prompt Injection Robustness27 Oct
- CAT-alyzing Honesty: Continuous Adversarial Training for Transferable LLM Honesty15 Oct
- Breaking the Tie: A Cluster-Aware Routing Framework for Large Language Models5 Oct
- Replay the Curvature: Accurate and Scalable NVFP4 Quantization for Large Language Model Inference29 Sep
- Shared Autoregressive Context Can Distort Relationships in Synthetic Data26 Sep
- Osprey: Target-agnostic Pre-training Makes Stronger Drafters in Speculative Decoding8 Sep
- The Illusion of Safety: Multi-Tier Verification of AI vs. Human C++ Code30 Jun
Same family
Llama 3- G
- 5.5
- 4.8
- 7.3
- 6.6
- 6.9
- 5.4
- 4.8
Want Llama 3.3 70B on OSMEXO?
Sign in to request it. Requests are reviewed; a request does not guarantee availability.