Cohere Models

AvalAI provides access to Cohere's suite of models designed for enterprise use cases, including advanced chat, retrieval-augmented generation (RAG), embedding, and reranking capabilities.

Chat Models (Command Series)

Cohere's Command models are optimized for conversational AI and long-context tasks.

Command R+

Cohere's most powerful and scalable model, excelling at complex RAG and tool use tasks.

Feature	Details
Model ID	command-r-plus
Context window	128,000 tokens
Capabilities	Chat, Advanced RAG, Tool Use (Function Calling), Multilingual
Input pricing	$0.50 / 1M tokens
Output pricing	$1.50 / 1M tokens
Strengths	State-of-the-art for RAG, complex workflows, enterprise-grade reliability
Best for	Demanding RAG applications, multi-step tool use, complex business logic

response = client.chat.completions.create(
    model="command-r-plus",
    messages=[
        {
            "role": "user",
            "content": "Based on the provided documents [cite documents], summarize the key findings regarding market trends.",
        },
        # Add documents/citations as needed by Cohere's chat endpoint structure
    ],
    # Add tool definitions if using Tool Use
)

Responses API version This version uses `gpt-5.5` because `command-r-plus` may not be enabled for `/v1/responses` in the current AvalAI model data.

Use this version when the selected model supports /v1/responses. messages moves to input, and the final text is read from response.output_text.

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["AVALAI_API_KEY"],
    base_url="https://api.avalai.ir/v1",
)

response = client.responses.create(
    model="gpt-5.5",
    input=[
        {
            "role": "user",
            "content": [
                {
                    "type": "input_text",
                    "text": "Based on the provided documents [cite documents], summarize the key findings regarding market trends.",
                },
                {"type": "input_file", "file_id": "file_abc123"},
            ],
        }
    ],
)

print(response.output_text)

• messages → input
• system message → instructions or a developer item
• choices[0].message.content → response.output_text
• for tools and multimodal output, inspect response.output by item type.

Command R

A highly capable and scalable model balancing performance and cost for RAG and tool use.

Feature	Details
Model ID	command-r
Context window	128,000 tokens
Capabilities	Chat, RAG, Tool Use (Function Calling), Multilingual
Input pricing	$0.50 / 1M tokens (Note: Typically lower than R+, check Cohere pricing)
Output pricing	$1.50 / 1M tokens (Note: Typically lower than R+, check Cohere pricing)
Strengths	Strong balance of performance and cost for RAG and tool use
Best for	General RAG applications, chatbots requiring citations, tool-enabled workflows

response = client.chat.completions.create(
    model="command-r",
    messages=[
        {"role": "user", "content": "What is Cohere's Command R model?"},
    ],
)

Responses API version This version uses `gpt-5.5` because `command-r` may not be enabled for `/v1/responses` in the current AvalAI model data.

Use this version when the selected model supports /v1/responses. messages moves to input, and the final text is read from response.output_text.

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["AVALAI_API_KEY"],
    base_url="https://api.avalai.ir/v1",
)

response = client.responses.create(
    model="gpt-5.5",
    instructions="You are a helpful assistant.",
    input="What is Cohere",
)

print(response.output_text)

• messages → input
• system message → instructions or a developer item
• choices[0].message.content → response.output_text
• for tools and multimodal output, inspect response.output by item type.

Note: Older models like command and cohere.command-light might also be available with smaller context windows.

Rerank Models

Cohere's Rerank models improve the quality of search results for RAG systems by re-ordering documents based on relevance to the query.

Rerank v4.0 Pro

The highest-quality reranking model with 32,768-token context window—8x larger than v3.5.

Feature	Details
Model ID	cohere-rerank-v4.0-pro
Context Window	32,768 tokens
Input type	Query string, list of document texts/IDs
Output type	Re-ordered list of documents with relevance scores [0, 1]
Languages	100+ languages supported
Pricing	$0.0025 / query
Strengths	Highest accuracy, long document support, multilingual
Best for	Production RAG systems requiring maximum relevance accuracy

import requests

response = requests.post(
    "https://api.avalai.ir/v1/rerank",
    headers={"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"},
    json={
        "model": "cohere-rerank-v4.0-pro",
        "query": "What is the capital of Canada?",
        "documents": [
            "Ottawa is the capital of Canada.",
            "Toronto is the largest city in Canada.",
        ],
        "top_n": 2,
    },
)

Rerank v4.0 Fast

A cost-effective reranking model optimized for high-throughput applications with minimal latency.

Feature	Details
Model ID	cohere-rerank-v4.0-fast
Context Window	32,768 tokens
Input type	Query string, list of document texts/IDs
Output type	Re-ordered list of documents with relevance scores [0, 1]
Languages	100+ languages supported
Pricing	$0.002 / query
Strengths	Fast, cost-effective, good accuracy
Best for	High-throughput applications with cost sensitivity

import requests

response = requests.post(
    "https://api.avalai.ir/v1/rerank",
    headers={"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"},
    json={
        "model": "cohere-rerank-v4.0-fast",
        "query": "What is the capital of Canada?",
        "documents": [
            "Ottawa is the capital of Canada.",
            "Toronto is the largest city in Canada.",
        ],
        "top_n": 2,
    },
)

Rerank v3.5

The previous generation reranking model.

Feature	Details
Model ID	cohere.rerank-v3-5:0
Input type	Query string, list of document texts/IDs
Output type	Re-ordered list of documents with relevance scores
Pricing	$1.00 / 1,000 Search Units
Strengths	Significantly improves search relevance for documents
Best for	Enhancing RAG systems, semantic search applications

# Example using Cohere's SDK (AvalAI might use a compatible endpoint)
# results = co.rerank(query="What is the capital of Canada?", documents=docs, model="cohere.rerank-v3-5:0")

Embedding Models (Embed Series)

Cohere's Embed models generate high-quality text and (for v3 Multilingual) image embeddings for semantic search and other representation tasks.

Embed English v3

State-of-the-art English embedding model.

Feature	Details
Model ID	cohere.embed-english-v3.0
Dimensions	1024
Max input tokens	512
Input pricing	$0.10 / 1M tokens
Strengths	Top performance on English embedding benchmarks
Best for	Semantic search, clustering, classification (English)

response = client.embeddings.create(
    model="cohere.embed-english-v3.0",
    input=["Your text here", "Another piece of text"],
    input_type="search_document",  # Or: search_query, classification, clustering
)

Embed Multilingual v3

State-of-the-art multilingual embedding model, supporting 100+ languages and image inputs.

Feature	Details
Model ID	cohere.embed-multilingual-v3.0
Dimensions	1024
Max input tokens	512 (text)
Input pricing	$0.10 / 1M tokens (text)
Strengths	High performance across many languages, supports text & image embeddings
Best for	Multilingual semantic search, cross-modal search (text/image)

# Text Embedding
response = client.embeddings.create(
    model="cohere.embed-multilingual-v3.0",
    input=["Aquí está tu texto", "Ein weiterer Text"],
    input_type="search_document",
)

# Image Embedding requires specific handling, check cohere.AvalAI docs

Embed v4 (AWS)

Cohere's latest multimodal embedding model supporting text, images, and mixed content including PDFs. Features a massive 128k context window and flexible output dimensions. Served via AWS infrastructure.

Feature	Details
Model ID	cohere.embed-v4:0
Provider	AWS
Dimensions	256, 512, 1024, or 1536 (default)
Max input tokens	128,000 tokens
Input types	Text, Images, Mixed texts/images (PDFs)
Text pricing	Contact for pricing
Image pricing	Contact for pricing
Similarity metrics	Cosine Similarity, Dot Product Similarity, Euclidean Distance
Strengths	Multimodal support, large context window, flexible dimensions
Best for	Document embeddings, PDF processing, multimodal search, RAG systems

# Text Embedding
response = client.embeddings.create(
    model="cohere.embed-v4:0",
    input=["Your text here", "Another piece of text"],
    encoding_format="float",
)

# You can also specify custom dimensions
response = client.embeddings.create(
    model="cohere.embed-v4:0",
    input=["Your text here"],
    dimensions=1024,  # Choose from 256, 512, 1024, or 1536
    encoding_format="float",
)

Embed v4 (Azure) - High Rate Limit

The same Cohere Embed v4 model served via Azure AI services infrastructure. Offers up to 30x higher rate limits compared to the AWS endpoint, ideal for high-throughput production workloads.

Feature	Details
Model ID	embed-v-4-0
Provider	Azure AI Services
Dimensions	256, 512, 1024, or 1536 (default)
Max input tokens	128,000 tokens
Input types	Text, Images, Mixed texts/images (PDFs)
Text pricing	$0.12 / 1M tokens
Image pricing	$0.47 / 1M tokens
Similarity metrics	Cosine Similarity, Dot Product Similarity, Euclidean Distance
Rate limits	Up to 30x higher than AWS endpoint
Strengths	High throughput, enterprise-grade stability, identical quality
Best for	High-throughput applications, production RAG systems, batch processing

# Text Embedding with high rate limits
response = client.embeddings.create(
    model="embed-v-4-0",
    input=["Your text here", "Another piece of text"],
    encoding_format="float",
)

# You can also specify custom dimensions
response = client.embeddings.create(
    model="embed-v-4-0",
    input=["Your text here"],
    dimensions=1024,  # Choose from 256, 512, 1024, or 1536
    encoding_format="float",
)

Note: Both cohere.embed-v4:0 (AWS) and embed-v-4-0 (Azure) produce identical embeddings. Choose based on your rate limit and throughput requirements.

Note: Light and v2 versions of embedding models might also be available.

Using Cohere Models via AvalAI

Access Cohere models using standard AvalAI API endpoints. Chat models use the OpenAI-compatible /chat/completions endpoint. Embedding and Rerank models may use /embeddings and potentially custom endpoints compatible with Cohere's API structure. Check AvalAI documentation for specifics on Rerank usage.

from openai import OpenAI

client = OpenAI(
    api_key="your-avalai-api-key",  # Replace with your actual API key
    base_url="https://api.avalai.ir/v1",  # AvalAI API endpoint
)

# Example using Command R+ Chat
response = client.chat.completions.create(
    model="cohere.command-r-plus",
    messages=[{"role": "user", "content": "What are Cohere's main product areas?"}],
)

# Example using Embed English v3
embed_response = client.embeddings.create(
    model="cohere.embed-english-v3.0", input=["This is a test sentence."]
)

print(response.choices[0].message.content)
# print(embed_response.data[0].embedding)

Responses API version This version uses `gpt-5.5` because `cohere.command-r-plus` may not be enabled for `/v1/responses` in the current AvalAI model data.

Use this version when the selected model supports /v1/responses. messages moves to input, and the final text is read from response.output_text.

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["AVALAI_API_KEY"],
    base_url="https://api.avalai.ir/v1",
)

response = client.responses.create(
    model="gpt-5.5",
    instructions="You are a helpful assistant.",
    input="What are Cohere",
)

print(response.output_text)

• messages → input
• system message → instructions or a developer item
• choices[0].message.content → response.output_text
• for tools and multimodal output, inspect response.output by item type.

Related Resources

• Chat Completions API
• Embeddings API
• Model Index
• Authentication
• Rate Limits