Inference
EN
Meta

Llama 3.3 70B

Llama 3.3 70B is a standard model from Meta. 14 benchmarks count toward its index, in 7 categories.

  • Open weights
  • Standard
  • Input: text
  • Released 2024-12
RankedEnough results for an index.
IndexAll categories, out of 100.28.2±5.3Median 51.0
CoverageIndex weight that has results.90%
SpeedOutput tokens per second.—Median 67/s
First tokenSeconds to the first output token.—Median 1.3 s
Input / 1MUS dollars per 1M input tokens.—Median $0.60
Output / 1MUS dollars per 1M output tokens.—Median $2.55
ContextMaximum tokens in one request.128KMedian 262K
EloLMArena rating and rank.—Median 1422

Run locally

70.55B parameters
16-bit
149.2 GB
8-bit
79.7 GB
4-bit
45.9 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy fits in 48 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.2 counted27.2205 / 218
CodingCode writing and repair.1 counted28.9—
ReasoningLogic problems and puzzles.2 counted27.5214 / 227
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.4 counted29.1241 / 268
MultilingualTasks in many languages.1 counted34.1—
InstructionTasks with strict prompt rules.2 counted40.944 / 63
MathMath problems.2 counted15.6164 / 168

Results

14 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Multilingual Grade School MathMultilingual91.1%2 / 13
Instruction-Following EvalInstructionAbove fitFar above this model's index.92.1%20 / 70
Evaluating Large Language Models Trained on CodeCoding88.4%8 / 22
Massive Multitask Language UnderstandingKnowledge86.0%16 / 34
Artificial Analysis IFBenchInstruction47.1%84 / 146
MATH level 5Math41.6% ±1.257 / 88
Critical Physics TasksReasoning0.0%150 / 214
GPQA diamond · Epoch AIKnowledge47.4% ±2.9155 / 217
GDPval-AA normalizedAgentic0.0%101 / 133
Massive Multitask Language Understanding ProfessionalKnowledge68.9%57 / 75
τ²-Bench Tool-Agent-User EvaluationAgentic26.6%116 / 149
OTIS Mock AIME 2024-2025Math5.1% ±2.4169 / 200
Graduate-Level Google-Proof Q&AKnowledge50.5%122 / 141
Artificial Analysis GPQA DiamondKnowledge49.8%174 / 199
Artificial Analysis Long Context ReasoningReasoning15.7%196 / 217
Artificial Analysis Humanity's Last ExamKnowledge3.6%210 / 222

Sources

BenchLMCC BY-NC 4.0Epoch AICC BY

History

5 releases
Index29.3 → 28.2
23 Oct 202612 Nov 2026

Questions

How good is Llama 3.3 70B?
Its index is 28.2 out of 100. That is place 255 of 283 ranked models.
What is Llama 3.3 70B best at?
Instruction: 40.9 out of 100, place 44 of 63.
What is the context window of Llama 3.3 70B?
128K tokens.
Is Llama 3.3 70B open source?
Yes. Its weights are published.
When was Llama 3.3 70B released?
December 2024.

Badge

Llama 3.3 70B on Inference360: place 255 of 283, index 28.2
HTML
<a href="https://inference360.ai/models/llama-3-3-70b"><img src="https://inference360.ai/badges/llama-3-3-70b.svg" alt="Llama 3.3 70B on Inference360: place 255 of 283, index 28.2" height="20"></a>
Markdown
[![Llama 3.3 70B on Inference360: place 255 of 283, index 28.2](https://inference360.ai/badges/llama-3-3-70b.svg)](https://inference360.ai/models/llama-3-3-70b)

Compare with

Devstral Small28.2Gemma 4 E4B28.2GPT-4o28.7Grok 229.1Llama 4 Maverick34.4

More from Meta

All models
Llama 3.3 70B InstructUnrankedLlama 3.2 1B InstructUnrankedLlama 3.2 3B InstructUnrankedLlama 4 Maverick34.4Llama 4 Scout26.8