Foundation Model · Amazon BedrockLlama FamilyLightweight TierTextMulti-Tier Pricing

Llama 3 8B

Amazon Bedrock token pricing and cost optimization reference for Llama 3 8B.Rates normalized to USD per 1 Million Tokens ($/1M) with prompt caching and batch API discounts across 7 AWS regions.

ProviderMeta

Maker HQMenlo Park, California, USA

Class / TierLightweight

ModalityText

Prompt CachingSave 75%

Batch APISave 50%

AWS Regions7 Global

Bedrock Model IDmeta.llama-3-8b-v1:0

Input / Prompt

$0.30/ 1M tokens

$0.0003 per 1K tokens

Output / Completion

$0.60/ 1M tokens

$0.0006 per 1K tokens

Prompt Cache Read

$0.07/ 1M tokens

Save 75% vs prompt

Batch Inference

$0.15/ 1M in

Save 50% via Batch API

Price Ratio

1 : 2.0

Prompt to completion factor

Regional Reach

7AWS Regions

Compare 7 regions ↓

Alternatives to Llama 3 8B

Option	Price vs Llama 3 8B	Why look at it
Llama 4 Scout 17B Compare →	input-43%-$0.13/1M output+10%+$0.06/1M Standard tier, per 1M tokens	Next step down in price from Meta
Llama 4 Maverick 17B Compare →	input-20%-$0.06/1M output+62%+$0.37/1M Standard tier, per 1M tokens	Next step up in price from Meta
Llama 3.3 70B Compare →	input+140%+$0.42/1M output+20%+$0.12/1M Standard tier, per 1M tokens	Same Llama line, different version (balanced tier)
Llama 3.1 70B Latency Optimized Compare →	input+200%+$0.60/1M output+50%+$0.30/1M Standard tier, per 1M tokens	Also from Meta (balanced tier)

Llama 3 8B Spend Estimator

Model inference spend by prompt volume, cache hit rate, and batch processing.

Prompt Tokens (Millions / month)

M tokens

Completion Tokens (Millions / month)

M tokens

Prompt Cache Hit Rate(75% discount on cached context)

30%

Batch API Mode50% discount on non-realtime async jobs via S3

Estimated Monthly Spend

$0.00

Fresh Prompts:$0.00

Cached Prompts:$0.00

Completions:$0.00

Active: Input: $0.30/1M · Output: $0.60/1M

Key Cost Levers & Architecture

Practical mechanisms to reduce inference costs with Llama 3 8B on AWS

Prompt CachingSave ~75%

Caches repetitive system instructions, schemas, and document prefixes in memory with 5-minute and extended 1-hour TTL tiers.

Read Rate: $0.07 / 1M tokens
Write Rate (5m TTL): $0.38 / 1M tokens
Ideal for: Multi-turn conversations, RAG document queries, structured schemas

Batch InferenceSave 50%

Asynchronous bulk processing submitted via Amazon S3 with a standard 24-hour turnaround SLA.

Batch Input: $0.15 / 1M tokens
Batch Output: $0.30 / 1M tokens
Ideal for: Offline classification, synthetic data generation, benchmark evals

Cross-Region & Tiers7 Regions

AWS Bedrock supports standard on-demand inference, priority reserved throughput, and cross-region routing.

Standard: Lowest latency, pay-per-token with zero commitments
Cross-Region Routing: Automatically burst traffic to available regional capacity
GovCloud: Dedicated isolated compliance partitions with +20% uplift

Regional Pricing Matrix

Live rates across all 7 available AWS cloud regions

Baseline: $0.30/1M inGovCloud: $0.50/1M in7 Global Regions

AWS Region	Prompt Input $/1M	Completion Output $/1M	Cache Read $/1M	Batch Input $/1M	Pricing Tier
us-east-1 US East (N. Virginia)	$0.30	$0.60	—	—	Baseline
us-west-2 US West (Oregon)	$0.30	$0.60	—	—	Baseline
us-gov-west-1 AWS GovCloud (US-West)	$0.30	$0.60	—	—	GovCloud
ap-south-1 Asia Pacific (Mumbai)	$0.36	$0.72	—	—	+20% vs Baseline
ca-central-1 Canada (Central)	$0.35	$0.69	—	—	+17% vs Baseline
eu-west-2 EU (London)	$0.39	$0.78	—	—	+30% vs Baseline
sa-east-1 South America (Sao Paulo)	$0.50	$1.01	—	—	+67% vs Baseline

Developer Quickstart

Invoke Llama 3 8B via Amazon Bedrock Converse API

Python (boto3) · Amazon Bedrock Runtime

import boto3

# Amazon Bedrock Converse API invocation
client = boto3.client("bedrock-runtime", region_name="us-east-1")

response = client.converse(
    modelId="meta.llama-3-8b-v1:0",
    messages=[
        {
            "role": "user",
            "content": [{"text": "Summarize key features and cost levers of Llama 3 8B."}]
        }
    ],
    inferenceConfig={
        "maxTokens": 1024,
        "temperature": 0.7
    }
)

output_text = response["output"]["message"]["content"][0]["text"]
usage = response["usage"]
print(f"Response: {output_text}")
print(f"Usage: {usage['inputTokens']} in, {usage['outputTokens']} out")