Reasoning

Pearl Llama 3.3 70B Instruct Turbo

High-performance 70B instruction model at 80% discounted pricing

About model

Pearl Llama 3.3 70B Instruct Turbo is Meta's 70B parameter instruction-tuned model running on Pearl's Proof of Useful Work protocol, delivering identical performance at 80% discounted pricing. Optimized with Together Turbo for high-throughput FP8 inference, the model supports 128K context, function calling, and multilingual instruction following. The Pearl kernel extracts cryptographic mining proofs during standard inference without affecting model quality or throughput.

Discounted Pricing

80%

Context Window

128K

Extended conversations and large codebase processing

Parameters

70B

High-throughput Turbo-optimized serving

Model key capabilities

80% Discounted Pricing: Powered by Pearl's Proof of Useful Work — mining rewards subsidize inference costs with zero impact on model quality or throughput
High-Throughput Turbo: Together Turbo FP8 optimization for low-latency, high-throughput serving of the 70B model
128K Context: Extended context for multi-turn conversations, large codebase processing, and document analysis
Instruction Following: Strong multilingual instruction tuning with function calling for conversational AI and agentic workflows

API usage

cURL
Python
Typescript

Endpoint:

meta-llama/pearl-Llama-3.3-70B-Instruct-Turbo

curl -X POST "https://api.together.xyz/v1/chat/completions" \
  -H "Authorization: Bearer $TOGETHER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "meta-llama/pearl-Llama-3.3-70B-Instruct-Turbo",
    "messages": [
      {
        "role": "user",
        "content": "What are some fun things to do in New York?"
      }
    ]
}'

from together import Together

client = Together()

response = client.chat.completions.create(
  model="meta-llama/pearl-Llama-3.3-70B-Instruct-Turbo",
  messages=[
    {
      "role": "user",
      "content": "What are some fun things to do in New York?"
    }
  ]
)
print(response.choices[0].message.content)

import Together from 'together-ai';
const together = new Together();

const completion = await together.chat.completions.create({
  model: 'meta-llama/pearl-Llama-3.3-70B-Instruct-Turbo',
  messages: [
    {
      role: 'user',
      content: 'What are some fun things to do in New York?'
     }
  ],
});

console.log(completion.choices[0].message.content);

Model card
Architecture Overview:
• 70B parameter dense transformer optimized with Together Turbo for high-throughput FP8 inference
• 128K token context window
• Instruction-tuned for conversational AI, coding, reasoning, and tool use
• Function calling support for agentic workflows

Pearl Proof of Useful Work:
• This endpoint runs on Pearl's Proof of Useful Work (PoUW) protocol, which extracts cryptographic mining proofs as a side effect of standard AI inference
• Model quality and throughput are preserved — the Pearl kernel operates at the matrix multiplication level without affecting model outputs
• Mining rewards flow back as a direct subsidy, enabling 80% discounted pricing
• Zero-knowledge proofs ensure no model weights or user data are exposed

Performance Characteristics:
• Identical performance to standard Llama 3.3 70B Instruct Turbo — all benchmarks and capabilities are preserved
• 81.9% MMLU for general knowledge
• Strong multilingual performance with instruction following across diverse tasks
• Together Turbo optimization for high-throughput, low-latency serving
‍
Prompting
Together AI API Access:
• Access Pearl Llama 3.3 70B Instruct Turbo via Together AI APIs using the endpoint meta-llama/pearl-Llama-3.3-70B-Instruct-Turbo
• Authenticate using your Together AI API key in request headers
• Supports function calling and structured output
• $0.176 per million input tokens / $0.176 per million output tokens (80% discount)
• Available on Together AI serverless infrastructure
‍
Applications & use cases
Conversational AI & Assistants:
• General-purpose instruction following for chatbots and virtual assistants
• Multi-turn conversation with 128K context for extended interactions
• Multilingual support across diverse user bases

Coding & Development:
• Code generation, explanation, and debugging across multiple languages
• Function calling for tool-augmented coding workflows
• 128K context for processing large codebases

Production Workloads:
• High-throughput serving via Together Turbo optimization
• Content generation, summarization, and analysis at scale
• Cost-efficient production deployment at 80% discounted pricing
‍

Related models

Model specifications

Model data

Model provider
Meta
Type
Reasoning
Main use cases
Reasoning
Features
Function Calling
Deployment
Serverless
Endpoint
meta-llama/pearl-Llama-3.3-70B-Instruct-Turbo
Parameters
70B
Context length
128K
Input price
$0.18 / 1M tokens
Output price
$0.18 / 1M tokens
Input modalities
Text
Output modalities
Text

Quantization level
FP8
Category
Chat

Run in Playground

Quickstart docs

Deploy model

Pearl Llama 3.3 70B Instruct Turbo

About model

API usage

Model card

Prompting

Applications & use cases