New Advanced Models: Gemini 3 Pro Preview and Kimi K2 Thinking

Date: 2025-11-18 / (1404-08-27)

Summary

We announce the addition of two flagship models: gemini-3-pro-preview from Google and kimi-k2-thinking from Moonshot AI. Gemini 3 Pro Preview is Google's most advanced model for complex reasoning tasks with 1M token context window, while Kimi K2 Thinking is Moonshot AI's newest flagship model with deep reasoning capabilities and multi-step tool use. Both models are available on v1/chat/completions with partial support on v1/responses.

---

Details

Google Gemini

Gemini 3 Pro Preview

Google's most advanced model in the Gemini series, gemini-3-pro-preview represents a significant advancement in AI capabilities. This natively multimodal model excels at complex reasoning tasks and can comprehend vast datasets from multiple information sources including text, audio, images, video, and entire code repositories.

Key Features:

• Context Window: 1M tokens for handling extensive conversations and documents
• Output Tokens: 64K tokens for comprehensive responses
• Advanced Capabilities: Native multimodal support (text, vision, audio), reasoning, function calling, structured outputs
• Architecture: Sparse mixture-of-experts (MoE) transformer-based model
• Knowledge Cutoff: January 2025
• Endpoint Support: Available on v1/chat/completions, partial support on v1/responses

Pricing Details:

Model	Input	Cached Input	Output	Special Pricing
gemini-3-pro-preview	$2.00/1M tokens	$0.825/1M tokens	$12.00/1M tokens	Above 200K: $4.00 input, $18.00 output

API Request/Response Examples

Example Request

curl https://api.avalai.ir/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "gemini-3-pro-preview",
    "messages": [
      {
        "role": "user",
        "content": "Analyze the technical architecture of a distributed system and propose optimization strategies."
      }
    ],
    "max_tokens": 2048
  }'

Example Response

{
  "id": "chatcmpl-abc123",
  "created": 1731945600,
  "model": "gemini-3-pro-preview",
  "object": "chat.completion",
  "system_fingerprint": null,
  "choices": [
    {
      "finish_reason": "stop",
      "index": 0,
      "message": {
        "content": "Let me analyze the distributed system architecture...\n\n## System Architecture Analysis\n\n1. **Load Balancing Layer**\n   - Current implementation uses round-robin...\n\n2. **Optimization Strategies**\n   - Implement adaptive load balancing...",
        "role": "assistant",
        "thinking_blocks": [],
        "annotations": []
      }
    }
  ],
  "usage": {
    "completion_tokens": 450,
    "prompt_tokens": 25,
    "total_tokens": 475,
    "completion_tokens_details": null,
    "prompt_tokens_details": {
      "audio_tokens": null,
      "cached_tokens": null,
      "text_tokens": 25,
      "image_tokens": null
    }
  },
  "estimated_cost": {
    "unit": "0.0054500000",
    "irt": 624.93,
    "exchange_rate": 114600
  }
}

SDK Usage Examples

Bash

curl https://api.avalai.ir/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "gemini-3-pro-preview",
    "messages": [
      {
        "role": "user",
        "content": "Solve this complex problem step by step."
      }
    ]
  }'

Python

from openai import OpenAI

client = OpenAI(api_key="your-avalai-api-key", base_url="https://api.avalai.ir/v1")

completion = client.chat.completions.create(
    model="gemini-3-pro-preview",
    messages=[
        {
            "role": "user",
            "content": "Solve this complex problem step by step.",
        }
    ],
)

print(completion.choices[0].message.content)

JavaScript

import { OpenAI } from "openai";

const client = new OpenAI({
  apiKey: process.env.AVALAI_API_KEY,
  baseURL: "https://api.avalai.ir/v1",
});

const completion = await client.chat.completions.create({
  model: "gemini-3-pro-preview",
  messages: [
    {
      role: "user",
      content: "Solve this complex problem step by step.",
    },
  ],
});

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

---

Moonshot AI

Kimi K2 Thinking

Moonshot AI's newest flagship model, kimi-k2-thinking, is a general-purpose agentic reasoning model designed for deep reasoning and multi-step tool use. This model excels at solving highly complex problems through extended reasoning chains and sequential tool calls.

Key Features:

• Deep Reasoning: Extended reasoning capabilities with reasoning_content field
• Multi-Step Tool Use: Designed to perform deep reasoning across multiple tool calls
• Agentic Performance: Excels at planning and executing complex multi-step tasks
• Advanced Problem Solving: Capable of tackling the hardest problems through step-by-step reasoning
• Endpoint Support: Available on v1/chat/completions, partial support on v1/responses

Pricing Details:

Model	Input	Cached Input	Output	Special Pricing
kimi-k2-thinking	$0.60/1M tokens	$0.15/1M tokens	$2.50/1M tokens	Search context: $0.005 per query

API Request/Response Examples

Example Request

curl https://api.avalai.ir/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "kimi-k2-thinking",
    "messages": [
      {
        "role": "user",
        "content": "Design a comprehensive digital marketing strategy for a tech startup"
      }
    ],
    "max_tokens": 16000,
    "temperature": 1.0,
    "stream": true
  }'

Example Response

{
  "id": "chatcmpl-xyz789",
  "created": 1731945600,
  "model": "kimi-k2-thinking",
  "object": "chat.completion",
  "system_fingerprint": null,
  "choices": [
    {
      "finish_reason": "stop",
      "index": 0,
      "message": {
        "content": "## Comprehensive Digital Marketing Strategy\n\n### Phase 1: Foundation\n1. **Brand Identity Development**\n   - Define core values and mission\n   - Create visual identity system\n\n### Phase 2: Channel Strategy\n...",
        "role": "assistant",
        "reasoning_content": "Let me think through this systematically. First, I need to understand the startup's context... Then I'll structure a multi-phase approach...",
        "thinking_blocks": [],
        "annotations": []
      }
    }
  ],
  "usage": {
    "completion_tokens": 1250,
    "prompt_tokens": 20,
    "total_tokens": 1270,
    "completion_tokens_details": null,
    "prompt_tokens_details": {
      "audio_tokens": null,
      "cached_tokens": null,
      "text_tokens": 20,
      "image_tokens": null
    }
  },
  "estimated_cost": {
    "unit": "0.0034125000",
    "irt": 391.07,
    "exchange_rate": 114600
  }
}

SDK Usage Examples

Bash

curl https://api.avalai.ir/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AVALAI_API_KEY" \
  -d '{
    "model": "kimi-k2-thinking",
    "messages": [
      {
        "role": "user",
        "content": "Solve this complex problem."
      }
    ],
    "max_tokens": 16000,
    "temperature": 1.0
  }'

Python

from openai import OpenAI

client = OpenAI(api_key="your-avalai-api-key", base_url="https://api.avalai.ir/v1")

completion = client.chat.completions.create(
    model="kimi-k2-thinking",
    messages=[
        {
            "role": "user",
            "content": "Solve this complex problem.",
        }
    ],
    max_tokens=16000,
    temperature=1.0,
)

# Access reasoning content if available
message = completion.choices[0].message
if hasattr(message, "reasoning_content"):
    reasoning = getattr(message, "reasoning_content")
    print("Reasoning:", reasoning)

print("Answer:", message.content)

JavaScript

import { OpenAI } from "openai";

const client = new OpenAI({
  apiKey: process.env.AVALAI_API_KEY,
  baseURL: "https://api.avalai.ir/v1",
});

const completion = await client.chat.completions.create({
  model: "kimi-k2-thinking",
  messages: [
    {
      role: "user",
      content: "Solve this complex problem.",
    },
  ],
  max_tokens: 16000,
  temperature: 1.0,
});

// Access reasoning content if available
const message = completion.choices[0].message;
if ("reasoning_content" in message) {
  console.log("Reasoning:", message.reasoning_content);
}

console.log("Answer:", message.content);

Recommended Settings for Kimi K2 Thinking

For optimal performance with kimi-k2-thinking:

• max_tokens: Set ≥ 16,000 to ensure full reasoning_content and final content can be returned
• temperature: Use 1.0 for best performance
• stream: Enable streaming (stream: true) for better user experience and to avoid timeout issues
• Multi-step reasoning: Include the entire reasoning_content from previous responses in your context

---

Google Model Deprecations

Google has deprecated the following models, effective immediately. We recommend migrating to gemini-3-pro-preview or other current models. See our deprecations page for more details.

Deprecated Models:

• gemini-2.0-flash-exp
• gemini-2.0-flash-lite-preview
• gemini-2.0-flash-lite-preview-02-05
• gemini-2.0-flash-thinking-exp
• gemini-2.0-flash-thinking-exp-01-21
• gemini-2.0-flash-thinking-exp-1219
• gemini-2.5-flash-lite-preview-06-17
• gemini-2.5-flash-preview-05-20
• gemini-2.5-pro-preview-06-05
• gemini-2.5-pro-preview-03-25
• gemini-2.5-pro-preview-05-06

---

Related Links

• Google Models Documentation
• Moonshot AI Models Documentation
• Chat Completions API Reference
• Responses API Reference
• Model Pricing
• Model Deprecations
• Reasoning Models Guide
• Function Calling Guide