> ## Documentation Index
> Fetch the complete documentation index at: https://docs.leapx-hub.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Changelog 🔥

> LeapX API latest updates, model releases, and pricing changes

Welcome to LeapX API changelog. Here you can find our latest model releases, pricing updates, and feature announcements.

<Tip>
  **Bookmark**: Press Ctrl+D to bookmark this page and stay updated!
</Tip>

## 🔥 Latest Updates

<CardGroup cols={1}>
  <Card title="12/12 GPT-5.2 Series" icon="rocket">
    OpenAI's newest flagship iteration! Better reasoning and coding, same price as GPT-5.1; GDPval real-world tasks: 70.9% on par with 14-year experts; ARC-AGI 2: 52.9%.
  </Card>

  <Card title="12/04 SeeDream 4.5" icon="image">
    BytePlus best 4K image generation model! 1.2B parameters, enhanced text rendering, supports 10 reference images, \$0.035/image!
  </Card>

  <Card title="11/25 Claude Opus 4.5" icon="brain">
    Anthropic's latest flagship! SWE-bench 80.9% beats all competitors, price reduced to 1/3, supports Claude Code!
  </Card>

  <Card title="11/20 Gemini 3 Pro Preview" icon="sparkles">
    Google's latest multimodal model! LMArena 1501 Elo #1, 1M context, thinking output support!
  </Card>

  <Card title="11/14 GPT-5.1 Series" icon="rocket">
    OpenAI's latest iteration! Intelligence and speed balanced, SWE-bench 76.3%, 24-hour caching!
  </Card>
</CardGroup>

## 📅 December 2025

### 🚀 GPT-5.2 Series Released (12/12)

OpenAI's year-end upgrade for Chat, Reasoning, and Code:

* **Performance**: Improved reasoning chain; more stable code/tool calls; better SWE-bench than 5.1
* **Pricing**: Same as GPT-5.1, keeping strong value
* **Compatibility**: Streaming, tool calling, and thinking/reasoning remain backward-compatible
* **Productivity focus**: Targeted at spreadsheets, decks, long docs, and coding; positioned as a "brainwork partner"
* **GDPval benchmark**: Across 44 professions and 1,320 real tasks, GPT-5.2 matches or beats 14-year experts in 70.9% of cases
* **ARC-AGI 2 leap**: Thinking version scores 52.9% (up from 17.6%), notably ahead of Gemini 3 Pro / Claude peers
* **Single mega-agent**: Lower tool-call latency, smoother multi-tool orchestration
* **Vision + long context**: Industrial-grade vision (e.g., circuit boards, wave simulations) with ultra-long context for multi-file and long-doc analysis

### 🎨 SeeDream 4.5 Released

BytePlus latest image generation model!

* **Ultra-high resolution**: 4K image generation
* **Parameter upgrade**: 1.2B parameters
* **Text rendering**: Significantly enhanced
* **Reference images**: Up to 10 reference images
* **Low price**: Only \$0.035/image

## 📅 November 2025

### 🚀 Claude Opus 4.5 Released

Anthropic's latest flagship model!

* **#1 Coding**: SWE-bench Verified 80.9%
* **Price drop**: 1/3 of previous Opus 4
* **Claude Code**: Full support
* **200K context**: Extended context window

### 🌟 Gemini 3 Pro Preview Released

Google's latest multimodal model!

* **LMArena #1**: 1501 Elo score
* **1M context**: Ultra-long context
* **Multimodal king**: Text/image/video/audio
* **Thinking output**: Full reasoning process

### 🚀 GPT-5.1 Series Released

OpenAI's latest iteration!

* **Performance**: SWE-bench Verified 76.3%
* **Speed**: 2-5x faster for simple tasks
* **Caching**: 24-hour prompt caching
* **Tools**: Apply\_Patch and Shell tools

## 📅 October 2025

### 🎥 Sora 2 Video Generation API

OpenAI's revolutionary video model!

* **Audio sync**: Auto-generated sound effects
* **No watermark**: Commercial-grade output
* **Low price**: Starting at \$0.15/generation

### ⚡ Claude Haiku 4.5 Released

High value coding model!

* **SWE-bench 73.3%**: Excellent coding
* **2x faster**: Improved response speed
* **1/3 cost**: Great value

### 💻 GLM-4.6 Released

Zhipu AI code and reasoning enhanced!

* **200K context**: Extended context
* **Code enhanced**: Better programming
* **Reasoning upgraded**: Stronger logic

## 📅 September 2025

### 🚀 Claude Sonnet 4.5 Released

World's best coding model!

* **SWE-bench 77.2%**: Best coding performance
* **Full upgrade**: All capabilities improved

### 🌐 Grok Web Search Models

Native web search!

* **Grok-4-all / Grok-3-all**: Built-in web search
* **Real-time info**: Latest web information

### ⚡ Grok-4-fast Series Released

Ultra-long context at lowest price!

* **200K Tokens**: Extended context
* **93%+ price drop**: Best value

## 📅 August 2025

### 🚀 GPT-5 Series Released

OpenAI's latest flagship!

* **Same price**: 20% off
* **Full series**: GPT-5, GPT-5-mini, etc.

### 🎯 OpenAI Open Source Models

OpenAI's first open source models!

* **gpt-oss-120b**: 120B parameter model
* **gpt-oss-20b**: 20B lightweight version
* **Local deployment**: Edge computing support

## 📊 Update Statistics

<AccordionGroup>
  <Accordion title="December 2025 Highlights">
    * 🚀 GPT-5.2: GDPval 70.9% matches/beats 14-year experts; ARC-AGI 2 at 52.9%; major gains in reasoning/coding/office workflows
    * 🎨 SeeDream 4.5: 4K image generation, better text rendering, \$0.035/image
  </Accordion>

  <Accordion title="November 2025 Highlights">
    * 🚀 Claude Opus 4.5: SWE-bench 80.9%, world's best coding model
    * 🌟 Gemini 3 Pro Preview: LMArena 1501 Elo #1 globally
    * 🚀 GPT-5.1 Series: Intelligence and speed balanced
    * 🎨 SeeDream 4.5: 4K image generation, \$0.035/image
  </Accordion>

  <Accordion title="October 2025 Highlights">
    * ⚡ Claude Haiku 4.5: High value coding model
    * 🎬 Sora 2: Video generation revolution
    * 💻 GLM-4.6: Code and reasoning enhanced
  </Accordion>

  <Accordion title="September 2025 Highlights">
    * 🚀 Claude Sonnet 4.5: World's best coding model
    * 🌐 Grok web models: Native web search
    * ⚡ Grok-4-fast: 200K context at lowest price
  </Accordion>
</AccordionGroup>

***

<Note>
  **Update frequency**: This page is updated regularly. Bookmark and check often for latest announcements.

  **Tip**: New models often have special launch promotions. Try them early!
</Note>