Amazon Web Services is the world's largest cloud provider with 200+ services across compute, storage, databases, ML, and networking. Dominates in enterprise with the broadest global region footprint and the deepest service catalog, but pricing complexity and egress fees add up at scale.
Llama 3.3 70B inference pricing
- Developer
- Meta
- Quality rank
- #30
- Elo
- 1290
- Context
- 128K
- Weights
- Open
- Lowest output
- $0.400
Lowest output
$0.400
Median
$0.790
Highest
$0.900
5 results
| Provider | Plan | Price | Regions | Visit | |||
|---|---|---|---|---|---|---|---|
|
|
Llama 3.3 70B | $0.400 | $0.230 | 128K |
$0.400
/M tokens
Input $0.230/1M tokens
Blended $0.273
Verified
|
Global | Visit → |
|
|
Bedrock Llama 3.3 70B | $0.720 | $0.720 | 128K |
$0.720
/M tokens
Input $0.720/1M tokens
Blended $0.720
Verified
|
Global | Visit → |
|
|
Llama 3.3 70B Versatile | $0.790 | $0.590 | 128K |
$0.790
/M tokens
Input $0.590/1M tokens
Blended $0.640
Verified
|
Global | Visit → |
|
|
Llama 3.3 70B Turbo | $0.880 | $0.880 | 128K |
$0.880
/M tokens
Input $0.880/1M tokens
Blended $0.880
Verified
|
Global | Visit → |
| Fireworks AI | Llama 3.3 70B | $0.900 | $0.900 | 128K |
$0.900
/M tokens
Input $0.900/1M tokens
Blended $0.900
Verified
|
Global | Visit → |
Providers serving this model
Fireworks AI specializes in high-throughput open-model inference powered by its custom FireAttention kernel, delivering token generation speeds that routinely beat other hosting platforms. With HIPAA compliance and a broad catalog spanning Llama, DeepSeek, Qwen, and Mistral models, it is built for latency-sensitive production applications at scale.
Together AI provides blazing-fast hosted inference for open-weight models including Llama 3.1 (8B through 405B), DeepSeek V3, Qwen 2.5, and Mistral - all at prices far below closed-model APIs. Its optimized serving infrastructure and free tier for experimentation make it the go-to platform for teams that prefer open models without self-hosting overhead.
Groq runs inference on custom LPU (Language Processing Unit) silicon rather than GPUs, delivering unmatched tokens-per-second throughput that can make even 70B models feel instant. With ultra-low pricing on Llama and DeepSeek models and a free tier for experimentation, it is the speed leader in the inference market.
DeepInfra offers rock-bottom priced hosted inference across a wide catalog of open-weight models, often undercutting competitors by 50-80%. With per-token billing as low as $0.03/M input on small models and aggressive pricing on DeepSeek V3 and Llama 70B, it is the cost champion for high-volume, budget-sensitive inference workloads.
Frequently asked questions
How much does Llama 3.3 70B cost per million tokens?
The lowest input price we track for Llama 3.3 70B is $0.230 per million tokens. Output tokens cost more; the table shows input, output, and blended pricing for every inference provider.
What is the cheapest Llama 3.3 70B API provider?
Sort the table by output or blended price to find the cheapest Llama 3.3 70B endpoint. Prices for the same model vary widely between providers, so the cheapest provider can be several times less than the most expensive.
Which providers serve the Llama 3.3 70B API?
Every provider with a published Llama 3.3 70B endpoint appears above, with input and output token pricing, context window, and throughput.
What is Llama 3.3 70B's context window?
Llama 3.3 70B supports a 128K context window. A larger context window lets you pass more tokens (documents, code, history) in a single request.
Is Llama 3.3 70B open weight or closed source?
Llama 3.3 70B is an open-weight model, so you can self-host it on any GPU provider, which usually beats managed API pricing at scale.
What is blended LLM cost?
Blended cost weights input and output token prices by a typical 3:1 ratio so you can rank providers by one number instead of comparing two prices separately.