impulse

Llama 3.2 3B Instruct

Open weights

Llama 3.2 3B is a 3-billion-parameter multilingual large language model, optimized for advanced natural language processing tasks like dialogue generation, reasoning, and summarization.

API id Llama-3.2-3B-InstructReleased 25 Sep 2024 (Sep 2024)
Input / 1M
$0.02
at DeepInfra
Output / 1M
$0.02
at DeepInfra
Context window
131K
tokens
Max output
131K
tokens per reply
Hosts
7
7 listings
Released
Sep 2024
25 Sep 2024

Pricing

List prices at DeepInfra, in US dollars per million tokens.

TierInput / 1MOutput / 1M
Standard$0.02$0.02

What it costs a month

At DeepInfra list prices, 30 days.

  • Chat assistant5,000 requests a day, 2K in / 0.5K out$7.50
  • RAG over documents2,000 requests a day, 8K in / 0.6K out$10.32
  • Coding agent1,000 requests a day, 40K in / 2K out$25.20
  • Overnight batch20,000 requests a day, 4K in / 0.4K out$52.80
Use your own numbers

Where to run it

7 hosts serve Llama 3.2 3B Instruct. The comparison is on blended price, 3 parts input to 1 part output.

HostIn / out per 1M
DeepInfraInference
Llama-3.2-3B-Instruct
$0.02 / $0.02
Novita AIInference
llama-3.2-3b-instruct
$0.03 / $0.05
+75% vs headline
Together AIInference
Llama-3.2-3B-Instruct
$0.06 / $0.06
+200% vs headline
Fireworks AIInference
llama-v3p2-3b-instruct
$0.10 / $0.10
+400% vs headline
llama-3.2-3b-instruct
$0.0509 / $0.335
+510% vs headline
llama-3-2-3b-instruct
$0.15 / $0.15
+650% vs headline
HyperbolicInference
Llama-3.2-3B-Instruct
$0.12 / $0.30
+725% vs headline

Capabilities

Core

  • Streaming
  • Function calling
  • JSON mode
  • Structured outputs

Multimodal

  • Vision
  • Image input
  • Audio
  • Image generation
  • Video

Advanced

  • Reasoning
  • Tool calling
  • MCP
  • Prompt caching
  • Batch API
  • Embeddings
Takes in
Text
Produces
Text
Open weights
License: Llama Community License
Runs in about 7.7 GB of GPU memory
Runtimes: Ollama, TGI, llama.cpp, vLLM

Cheaper alternatives

Current models that cost less and keep most of what this one can do: tools, vision, reasoning, structured output, caching and enough context.

Same model, cheaper host

No host sells it for less than DeepInfra.

Smaller Meta models

No cheaper Meta model keeps its key capabilities.

Other frontier labs

No cheaper match from another frontier lab.

Open-weight picks

No cheaper open-weight match.

Benchmarks

Scores published on the model card.

  • BBH24.1%
  • GPQA3.8%
  • IFEval73.9%
  • MATH Lvl 517.7%
  • MMLU-Pro24.4%
  • MUSR1.4%
  • Open LLM Average24.2%

Model line

Newer and older versions, at each lab's own price.

  1. Sep 2024Llama 3.2 3B InstructLatest$0.02 / $0.02
  2. Sep 2024Llama3.2 3B Instruct$0.015 / $0.025

Call it

Using DeepInfra's API.

Python
from litellm import completion

response = completion(
    model="deepinfra/Llama-3.2-3B-Instruct",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

Quick answers

How much does Llama 3.2 3B Instruct cost?

On DeepInfra, Llama 3.2 3B Instruct costs $0.02 per million input tokens and $0.02 per million output tokens.

Which APIs offer Llama 3.2 3B Instruct?

Llama 3.2 3B Instruct runs on 7 APIs: DeepInfra, Novita AI, Together AI, Fireworks AI, Cloudflare Workers AI, IBM watsonx.ai and Hyperbolic.

What is Llama 3.2 3B Instruct's context window?

131K tokens, with up to 131K tokens in one reply.

Is Llama 3.2 3B Instruct being retired?

No retirement date is scheduled for it on any API we track.

Is there a newer version of Llama 3.2 3B Instruct?

No. Llama 3.2 3B Instruct is the newest model in its line.

When was Llama 3.2 3B Instruct released?

Meta released it on 25 Sep 2024.

What is Llama 3.2 3B Instruct's API model id?

Use Llama-3.2-3B-Instruct with DeepInfra. Other hosts use their own ids: llama-3.2-3b-instruct on Novita AI, llama-v3p2-3b-instruct on Fireworks AI, llama-3.2-3b-instruct on Cloudflare Workers AI, llama-3-2-3b-instruct on IBM watsonx.ai.

Sources

Raw catalog entry for Llama-3.2-3B-Instruct on DeepInfra
{
  "max_input_tokens": 131072,
  "max_output_tokens": 131072,
  "input_cost_per_token": 2e-8,
  "output_cost_per_token": 2e-8,
  "litellm_provider": "deepinfra",
  "mode": "chat",
  "supports_tool_choice": true,
  "supports_function_calling": true
}