Llama 3.1 8B
Meta · dense · 8.03B parameters · 131,072 context
Parameters
8.03B
Context Window
128K tokens
Architecture
Dense
Best GPU
A30
Cheapest API
$0.00/M
Quality Score
58/100
Intelligence Brief
Llama 3.1 8B is a 8.03B parameter DENSE model from Meta, featuring Grouped Query Attention (GQA) with 32 layers and 4,096 hidden dimensions. With a 131,072 token context window, it supports tools, structured output, code, math, multilingual. On standardized benchmarks, it achieves MMLU 69.4, HumanEval 40.2, GSM8K 79.6. The most cost-effective API deployment is via scaleway at $0.00/M output tokens. For self-hosted inference, A30 delivers optimal throughput at $332/month.
Provider pricing
10 providers · canonical: together| Provider | Input $/M | Output $/M ▲ | Notes |
|---|---|---|---|
| scaleway | free | free | cheapest input · cheapest output |
| featherless | free | free | cheapest input · cheapest output |
| novita | $0.020 | $0.050 | — |
| openrouter | $0.020 | $0.050 | — |
| groq | $0.050 | $0.080 | — |
| deepinfra | $0.050 | $0.080 | — |
| cerebras | $0.100 | $0.100 | — |
| togethercanonical | $0.180 | $0.180 | — |
| together-ai | $0.180 | $0.180 | — |
| fireworks | $0.200 | $0.200 | — |
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 3.1 405BLlama 3.1 · 405B$1.00/M out
- Llama 3.1 70BLlama 3.1 · 70.6Bfree/M out
- HelpSteer2 Llama 3.1 70BLlama 3.1 · 70.6B$0.400/M out
- Llama 3.1 Nemotron 51BLlama 3.1 · 51B$0.400/M out
- Llama 3.1 Nemotron 70B InstructLlama 3.1 · 70.6B$0.880/M out
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
16.1 GB
FP8 Weights
8.0 GB
INT4 Weights
4.0 GB
GPU Compatibility Matrix
Llama 3.1 8B is compatible with 90% of GPU configurations across 41 GPUs at 3 precision levels.
GPU Recommendations
BF16 · 1 GPU · vllm
100/100
score
Throughput
313.7 tok/s
Latency (ITL)
3.2ms
Est. TTFT
1ms
Cost/Month
$332
Cost/M Tokens
$0.40
BF16 · 1 GPU · vllm
100/100
score
Throughput
338.9 tok/s
Latency (ITL)
3.0ms
Est. TTFT
1ms
Cost/Month
$370
Cost/M Tokens
$0.42
BF16 · 1 GPU · vllm
100/100
score
Throughput
314.7 tok/s
Latency (ITL)
3.2ms
Est. TTFT
1ms
Cost/Month
$180
Cost/M Tokens
$0.22
Deployment Options
API Deployment
scaleway
$0.00/M
output tokens
Single GPU
A30
$332/mo
Min VRAM: 8 GB
Multi-GPU
RTX 3060 x2
189.8 tok/s
TP· $114/mo
API Pricing Comparison
| Provider | Input $/M | Output $/M | Badges |
|---|---|---|---|
| scaleway | $0.00 | $0.00 | Cheapest |
| featherless | $0.00 | $0.00 | |
| novita | $0.02 | $0.05 | |
| openrouter | $0.02 | $0.05 | |
| groq | $0.05 | $0.08 | |
| deepinfra | $0.05 | $0.08 | |
| cerebras | $0.10 | $0.10 | |
| together | $0.18 | $0.18 | |
| together-ai | $0.18 | $0.18 | |
| fireworks | $0.20 | $0.20 |
Cost Analysis
| Provider | Input $/M | Output $/M | ~Monthly Cost |
|---|---|---|---|
| scalewayBest Value | $0.00 | $0.00 | $0 |
| featherless | $0.00 | $0.00 | $0 |
| novita | $0.02 | $0.05 | $0 |
| openrouter | $0.02 | $0.05 | $0 |
| groq | $0.05 | $0.08 | $1 |
| deepinfra | $0.05 | $0.08 | $1 |
| cerebras | $0.10 | $0.10 | $1 |
| together | $0.18 | $0.18 | $2 |
| together-ai | $0.18 | $0.18 | $2 |
| fireworks | $0.20 | $0.20 | $2 |
Cost per 1,000 Requests
Short (500 tok)
$0.00
via scaleway
Medium (2K tok)
$0.00
via scaleway
Long (8K tok)
$0.00
via scaleway
Performance Estimates
Throughput by GPU
VRAM Breakdown (A30, BF16)
Precision Impact
bf16
16.1 GB
weights/GPU
~313.7 tok/s
fp8
8.0 GB
weights/GPU
int4
4.0 GB
weights/GPU
Quality Benchmarks
Capabilities
Features
Supported Frameworks
Supported Precisions
Where to Deploy Llama 3.1 8B
Self-Hosted Infrastructure
Similar Models
Hermes 3 8B
8.03B params · dense
Quality: 50
from $0.18/M
Aya 23 8B
8B params · dense
Quality: 50
from $0.60/M
DeepSeek R1 Distill 8B
8B params · dense
Quality: 88
from $0.20/M
InternLM3 8B
8B params · dense
Quality: 50
Llama 3 8B
8B params · dense
Quality: 63
from $0.04/M
Frequently Asked Questions
How much VRAM does Llama 3.1 8B need for inference?
Llama 3.1 8B requires approximately 16.1 GB of VRAM at BF16 precision, 8.0 GB at FP8, or 4.0 GB at INT4 quantization. Additional VRAM is needed for KV-cache (131072 bytes per token) and activations (~1.00 GB).
What is the best GPU for Llama 3.1 8B?
The top recommended GPU for Llama 3.1 8B is the A30 using BF16 precision. It achieves approximately 313.7 tokens/sec at an estimated cost of $332/month ($0.40/M tokens). Score: 100/100.
How much does Llama 3.1 8B inference cost?
Llama 3.1 8B API inference starts from $0.00/M input tokens and $0.00/M output tokens. Self-hosted inference costs depend on your GPU configuration — use our ROI calculator for a detailed breakdown.