Phind CodeLlama 34B V2 on DeepInfra

Name: Phind CodeLlama 34B V2 on DeepInfra
Brand: Phind
Price: 0.20 USD

ServerlessOpen Source

Last refreshed 2026-04-24. Next refresh: weekly.

Why use Phind CodeLlama 34B V2 on DeepInfra?

DeepInfra offers Phind CodeLlama 34B V2 with pay-as-you-go pricing at $0.20/1M input tokens. DeepInfra is a cloud inference platform offering cost-effective access to open-source AI models.

Compare Phind CodeLlama 34B V2 across 3 providers to find the best fit for your use case

Input / 1M

$0.20

Output / 1M

$0.45

Cache

Not sourced

Batch

Not sourced

Setup recipe

Python + curl

Install

pip install openai

Auth

export DEEPINFRA_API_KEY=...

Call

import os
from openai import OpenAI
client = OpenAI(
    api_key=os.environ["DEEPINFRA_API_KEY"],

Model ID

phind-codellama-34b-v2

Request example

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["DEEPINFRA_API_KEY"],
    base_url="https://api.deepinfra.com/v1/openai"
)
response = client.chat.completions.create(
    model="phind-codellama-34b-v2",
    messages=[{"role": "user", "content": "Hello"}]
)
print(response.choices[0].message.content)

Gotchas

DeepInfra uses "organization/model-name" format, e.g. "meta-llama/Meta-Llama-3-8B-Instruct" or "mistralai/Mistral-7B-Instruct-v0.3". See the DeepInfra model catalog for exact IDs.
The examples expect DEEPINFRA_API_KEY; rename it only if your application config maps the new variable.

Compare Phind CodeLlama 34B V2 Across Providers

Provider	Input (per 1M)	Output (per 1M)
DeepInfra	$0.20	$0.45
Together AI	$0.80	$0.80
Fireworks AI	$0.90	$0.90

Pricing

Type	Price (per 1M)
Input tokens	$0.20
Output tokens	$0.45

Capabilities

Structured Outputs

About Phind CodeLlama 34B V2

Phind CodeLlama 34B v2 is a large language model designed specifically for code generation tasks, built on the CodeLlama architecture. It generates high-quality code in multiple programming languages such as Python, C/C++, TypeScript, and Java, and is instruction-tuned for enhanced usability. The model demonstrates strong benchmark performance, achieving a 73.8% pass@1 score on the HumanEval benchmark. It has been fine-tuned on a proprietary dataset of 1.5 billion tokens, focusing on instruction-answer pairs. Additionally, it showcases multi-lingual capabilities beyond programming languages and offers various quantized versions like GPTQ and GGUF for optimized performance and reduced memory usage. Despite its impressive features, thorough testing is advised prior to real-world deployment, as its testing has been limited so far 1 2 7.