Llama 3.2 3B
Meta · dense · 3.21B parameters · 131,072 context
Parameters
3.21B
Context Window
128K tokens
Architecture
Dense
Best GPU
RTX 4070 Ti
Cheapest API
$0.06/M
Quality Score
55/100
Intelligence Brief
Llama 3.2 3B is a 3.21B parameter DENSE model from Meta, featuring Grouped Query Attention (GQA) with 28 layers and 3,072 hidden dimensions. With a 131,072 token context window, it supports tools, structured output, code, math, multilingual. On standardized benchmarks, it achieves MMLU 63.4, HumanEval 33, GSM8K 68. The most cost-effective API deployment is via together at $0.06/M output tokens. For self-hosted inference, RTX 4070 Ti delivers optimal throughput at $237/month.
Provider pricing
3 providers · canonical: together| Provider | Input $/M | Output $/M ▲ | Notes |
|---|---|---|---|
| togethercanonical | $0.060 | $0.060 | cheapest output |
| fireworks | $0.100 | $0.100 | — |
| openrouter | $0.051 | $0.340 | cheapest input |
Prices update via the nightly pricing cron + admin approvals at /admin/ingest-queue. The leaderboard's Input/Output cells show the canonical rate above; this table shows the full spread.
Recent changes
Loading…
Related models
5 suggestions- Llama 3.2 11B VisionLlama 3.2 · 11B$0.180/M out
- Llama 3.2 1BLlama 3.2 · 1.24Bfree/M out
- Llama 3.2 90B Vision InstructLlama 3.2 · 88.8B$1.20/M out
- Llama 3.2 90B VisionLlama 3.2 · 90B$0.900/M out
- SeamlessM4T v2 LargeSeamlessM4T · 2.3B—
Picks: same family first, then same vendor within ±2× params, then top tag-overlap matches. Price shown is the cheapest Output $/M across providers — the row's page shows the canonical anchor.
Architecture Details
Memory Requirements
BF16 Weights
6.4 GB
FP8 Weights
3.2 GB
INT4 Weights
1.6 GB
GPU Compatibility Matrix
Llama 3.2 3B is compatible with 100% of GPU configurations across 41 GPUs at 3 precision levels.
GPU Recommendations
BF16 · 1 GPU · vllm
100/100
score
Throughput
423.9 tok/s
Latency (ITL)
2.4ms
Est. TTFT
0ms
Cost/Month
$237
Cost/M Tokens
$0.21
BF16 · 1 GPU · vllm
100/100
score
Throughput
639.2 tok/s
Latency (ITL)
1.6ms
Est. TTFT
0ms
Cost/Month
$133
Cost/M Tokens
$0.08
BF16 · 1 GPU · vllm
100/100
score
Throughput
423.9 tok/s
Latency (ITL)
2.4ms
Est. TTFT
0ms
Cost/Month
$209
Cost/M Tokens
$0.19
Deployment Options
API Deployment
together
$0.06/M
output tokens
Single GPU
RTX 4070 Ti
$237/mo
Min VRAM: 3 GB
Multi-GPU
RTX 4070 Ti
423.9 tok/s
Best available config
API Pricing Comparison
| Provider | Input $/M | Output $/M | Badges |
|---|---|---|---|
| together | $0.06 | $0.06 | Cheapest |
| fireworks | $0.10 | $0.10 | |
| openrouter | $0.05 | $0.34 | Low Input |
Cost Analysis
| Provider | Input $/M | Output $/M | ~Monthly Cost |
|---|---|---|---|
| togetherBest Value | $0.06 | $0.06 | $1 |
| fireworks | $0.10 | $0.10 | $1 |
| openrouter | $0.05 | $0.34 | $2 |
Cost per 1,000 Requests
Short (500 tok)
$0.04
via together
Medium (2K tok)
$0.17
via together
Long (8K tok)
$0.60
via together
Performance Estimates
Throughput by GPU
VRAM Breakdown (RTX 4070 Ti, BF16)
Precision Impact
bf16
6.4 GB
weights/GPU
~423.9 tok/s
fp8
3.2 GB
weights/GPU
int4
1.6 GB
weights/GPU
Quality Benchmarks
Capabilities
Features
Supported Frameworks
Supported Precisions
Where to Deploy Llama 3.2 3B
Self-Hosted Infrastructure
Similar Models
Llama 3.2 1B
1.24B params · dense
Quality: 38
from $0.00/M
Replit Code v1.5 3B
3.3B params · dense
Quality: 50
Qwen 2.5 Coder 3B
3.1B params · dense
Quality: 50
Qwen 2.5 3B
3.09B params · dense
Quality: 58
from $0.10/M
StarCoder2 3B
3.03B params · dense
Quality: 29
from $0.10/M
Frequently Asked Questions
How much VRAM does Llama 3.2 3B need for inference?
Llama 3.2 3B requires approximately 6.4 GB of VRAM at BF16 precision, 3.2 GB at FP8, or 1.6 GB at INT4 quantization. Additional VRAM is needed for KV-cache (114688 bytes per token) and activations (~0.50 GB).
What is the best GPU for Llama 3.2 3B?
The top recommended GPU for Llama 3.2 3B is the RTX 4070 Ti using BF16 precision. It achieves approximately 423.9 tokens/sec at an estimated cost of $237/month ($0.21/M tokens). Score: 100/100.
How much does Llama 3.2 3B inference cost?
Llama 3.2 3B API inference starts from $0.06/M input tokens and $0.06/M output tokens. Self-hosted inference costs depend on your GPU configuration — use our ROI calculator for a detailed breakdown.