Subscribe to Updates: Follow our announcements to get model launches and promotional information first, don’t miss any important updates!
🔥 Latest Updates
Recent important updates:- 10/12 Gemini 2.5 Flash Latest Version gemini-2.5-flash-preview-09-2025 is live! Latest dated version!
- 10/03 Gemini Image Official Release Nano Banana upgraded! 10 aspect ratios, custom resolution support!
- 10/01 Sora 2 Video Generation OpenAI’s revolutionary video model is live! Audio-video sync, from $0.15/call, no watermark!
- 10/01 GLM-4.6 Released Zhipu AI latest version, 200K context, comprehensive code and reasoning improvements!
- 09/30 Claude Sonnet 4.5 World’s strongest coding model is live! SWE-bench 77.2%, supports Claude Code!
- 09/29 DeepSeek-V3.2-exp Latest experimental version launched early, first to integrate, priced below official!
- 09/26 GLM-4.5 Series Three versions fully launched, priced lower than Zhipu official, extremely cost-effective!
- 09/25 Grok Web Search Models grok-4-all and grok-3-all launched, native web search capability, no tool calling needed!
- 09/24 Grok-4-fast Series Ultra-long context 200K tokens, reasoning/non-reasoning dual modes, record low pricing!
- 09/24 Grok Code Fast 1 Code-specialized model launched, SWE-Bench 70.8%, performance champion!
- 09/16 OpenAI Codex Series gpt-5-codex series launched, optimized for programming, supports dual billing modes!
- 09/11 SeeDream 4.0 API seedream-4-0-250828 launched, BytePlus Volcano Engine strategic partnership, 35% off official price!
- 08/27 Gemini Image Generation gemini-2.5-flash-image-preview launched, Google’s strongest image model, currently popular, APIYI 50% off!
- 08/26 DeepSeek V3.1 Mixed reasoning mode launched, supports Think/Non-Think dual modes
- 08/25 Kimi K2 Official Release kimi-k2-250711 integrated, official Volcano Engine partnership version, stable and reliable
- 08/08 GPT-5 Full Series Officially launched, same price as official, ~20% off with top-up bonuses + exchange rate benefits
- 08/07 Claude Opus 4.1 Performance upgrade launched, more for the same price
- 08/06 OpenAI Open Source Models gpt-oss-120b and gpt-oss-20b launched, priced lower than DeepSeek R1/V3
- GPT-4.5-Preview Model discontinued, please switch to GPT-4.1
- Kimi K2 Model launched, 30% off official price
- Claude Full series 20% price reduction
- O3 Model 80% price reduction, biggest drop ever
📅 October 2025
✨ Gemini 2.5 Flash Latest Dated Version Launch
October 12, 2025Gemini 2.5 Flash Preview 09-2025 - Google’s Latest Version
-
🚀 New Model:
gemini-2.5-flash-preview-09-2025 -
📅 Version Notes:
- Latest dated version of Gemini 2.5 Flash (September 2025 release)
- Google’s official latest iteration of Gemini 2.5 Flash series
- Includes latest performance optimizations and feature improvements
-
🌟 Core Features:
- Continues the powerful capabilities of Gemini 2.5 Flash series
- Supports ultra-long context processing
- Multimodal understanding (text, images, video)
- Google’s latest technical optimizations and performance enhancements
-
💰 Pricing Advantage:
- APIYI offers highly competitive pricing
- Even better with top-up bonus promotions
- Official latest version with stronger performance
-
🛠️ Calling Method:
- Use standard chat completion endpoint:
/v1/chat/completions - Fully compatible with OpenAI API format
- Simply replace model name to use
- Use standard chat completion endpoint:
gemini-2.5-flash-preview-09-2025 is Google’s officially released latest dated version of Gemini 2.5 Flash. Recommend using the latest version for best performance and experience. More info: Google Official Docs
🍌 Gemini 2.5 Flash Image Official Release
October 3, 2025Gemini 2.5 Flash Image - Google Image Generation Official Release
-
🚀 Model Update:
- New model name:
gemini-2.5-flash-image - Old model name:
gemini-2.5-flash-image-preview(still available)
- New model name:
-
💰 Pricing Maintained:
- APIYI pricing: 0.039)
- Save about 36% compared to official
- Continues cost-effectiveness advantage
-
🌟 Core Upgrades:
- 10 Aspect Ratio Support: From cinematic landscape to social media portrait
- Custom Resolution: Specify output resolution
- Production Ready: Upgraded from preview to official release
- Image-only Output: Supports pure image output mode
-
🎨 Powerful Capabilities:
- Seamlessly blend multiple images
- Maintain character consistency, enrich storytelling
- Natural language precise editing
- Combined with Gemini’s rich world knowledge
-
🛠️ Calling Method:
- Recommended new model name:
gemini-2.5-flash-image - Old model name
gemini-2.5-flash-image-previewstill available - Other calling methods remain unchanged
- Fully compatible with existing code
- Recommended new model name:
New version supports 10 aspect ratios and custom resolution features. Recommend using
gemini-2.5-flash-image for best experience. Old model name still available.🎬 Sora 2 Video Generation API Launch
October 1, 2025 EveningSora 2 - OpenAI’s Revolutionary Video Generation Model
-
🚀 New Models:
sora_video2- Portrait video (704×1280)sora_video2-landscape- Landscape video (1280×704)sora-2-pro-all- Pro HD version (1024×1792)
-
💰 Great Pricing:
- Standard version: $0.15/call (10 second video)
- Pro version: $0.20/call (supports 15 seconds, HD)
- With top-up bonuses, from ~¥0.8/call
-
🎥 Core Features:
- Audio-Video Sync: Industry’s first audio-video synchronized generation
- Physical Reality Enhancement: Significantly improved video realism
- Watermark-free Output: Generated videos have no watermark (official has watermarks)
- Long Video Support: Up to 16 seconds of coherent narrative
-
🛠️ Generation Methods:
- Text-to-Video: Generate videos from pure text descriptions
- Image-to-Video: Supports 1 image reference for generation
- Supports URL and Base64 image upload
- Complete streaming output progress feedback
-
🌟 Technical Highlights:
- Endpoint:
/v1/chat/completions - Standard OpenAI API format
- Supports streaming output to view progress
- No invitation code needed, use immediately
- Endpoint:
Sora 2 is OpenAI’s revolutionary video generation model released on October 1, positioned as the “GPT-3.5 moment” for video generation. APIYI integrated official API first, priced from only $0.15/call, no invitation code required. Detailed docs: Sora 2 User Guide
💻 GLM-4.6 Strong Release
October 1, 2025GLM-4.6 - Zhipu AI Code and Reasoning Enhanced Edition
-
🚀 New Model:
glm-4.6 -
💰 Pricing Advantage:
- Input price: $0.50 / 1M tokens
- Output price: $1.75 / 1M tokens
- Extremely cost-effective, only 1/7 of Claude’s price
-
🧠 Core Upgrades:
- 200K Ultra-long Context: Expanded from 128K to 200K tokens
- Code Capability Enhancement: Excellent performance in Claude Code, Cline, Roo Code IDEs
- Advanced Reasoning: Significantly improved reasoning with tool use during reasoning
- Agent Enhancement: Stronger performance in tool calling and search-based agents
-
🌟 Performance Highlights:
- Programming: 95.0% accuracy, ranked 95th percentile
- Math: 96.0% accuracy, ranked 98th percentile
- Close to Claude Sonnet 4 (48.6% win rate)
- Significantly better than open-source baseline models
-
⚡ Writing Optimization:
- More human-preferred style and readability
- More natural role-playing scenarios
- Better visual effects for frontend page generation
GLM-4.6 is the latest version released by Zhipu AI (now rebranded as Z.AI) on September 30, with significant improvements in code generation, reasoning capabilities, and agent framework integration compared to GLM-4.5.
📅 September 2025
🚀 Claude Sonnet 4.5 Spectacular Launch
September 30, 2025Claude Sonnet 4.5 - Anthropic’s Strongest Coding Model
-
🚀 New Models:
claude-sonnet-4-5-20250929- Standard versionclaude-sonnet-4-5-20250929-thinking- Reasoning mode
-
💰 Pricing Advantage:
- Completely aligned with Anthropic official pricing
- Enjoy top-up bonus promotions for $100+ recharges
- With bonus discounts, about 20% off official pricing
- Pricing: 15 per 1M tokens (input/output)
-
🏆 World-class Coding Capability:
- SWE-bench Verified 77.2%: World’s strongest coding model
- OSWorld 61.4%: Leading in real computer task testing
- Supports Claude Code, excellent development experience
- Top choice for building complex agents
-
🧠 Core Upgrades:
- Autonomous 30-hour Operation: Massive improvement over Opus 4’s 7 hours
- Significantly Enhanced Reasoning and Math
- Supports sustained focus on complex multi-step tasks
- Comprehensive safety training upgrade
-
🌟 Real-world Applications:
- Devin planning performance improved 18%
- End-to-end evaluation score improved 12%
- Ideal for building autonomous software development systems
- Enterprise-grade complex business process automation
-
🛠️ Calling Methods:
- Standard version: Regular conversation and coding tasks
- Thinking mode: Complex problems requiring deep reasoning
- Fully compatible with Anthropic API format
Claude Sonnet 4.5 is Anthropic’s strongest AI model released on September 29, hailed as the “world’s best coding model”, capable of autonomously running for up to 30 hours handling complex tasks. APIYI same price as official, enjoy 20% off with top-ups.
🧠 DeepSeek-V3.2-exp Early Launch
September 29, 2025DeepSeek-V3.2-exp - DeepSeek’s Latest Experimental Version
-
🚀 New Model:
deepseek-v3.2-exp -
💰 Pricing Advantage:
- Input price: 0.28)
- Output price: 0.42)
- Save about 7% compared to official
- Better pricing in CNY (official ¥2/¥3)
-
🌟 Core Features:
- Official direct connection version, stable and reliable
- Launched first to meet customer needs
- DeepSeek’s latest experimental capabilities
- Continuous iteration and optimization
-
🛠️ Calling Method:
- Use standard Chat endpoint:
/v1/chat/completions - Compatible with OpenAI API format
- Pay-as-you-go, simple and transparent
- Use standard Chat endpoint:
DeepSeek-V3.2-exp is DeepSeek’s latest experimental version released today. APIYI integrated the official direct connection version first, providing service at prices slightly lower than official.
🤖 GLM-4.5 Series Models Launch
September 26, 2025GLM-4.5 - Zhipu AI High-Performance Language Model
-
🚀 New Models:
glm-4.5- Standard versionglm-4.5-air- Lightweight versionglm-4.5v- Multimodal version
-
💰 Pricing Advantage:
- All models priced lower than Zhipu official
- Better cost-effectiveness while ensuring stability
- Pay-as-you-go billing
-
🌟 Core Features:
- Comprehensive coverage of different performance scenarios
- Excellent Chinese understanding and generation
- Stable and reliable API service
- Compatible with standard OpenAI API format
GLM-4.5 series is Zhipu AI’s high-performance language model. APIYI offers pricing lower than official, continuously providing better choices in model coverage and cost-effectiveness.
🌐 Grok Web Search Models Official Launch
September 25, 2025Grok-4-all / Grok-3-all - Smart Models with Built-in Web Search
-
🚀 New Models:
grok-4-all- Grok-4 web search enhanced versiongrok-3-all- Grok-3 web search enhanced version
-
💰 Unified Pricing:
- Input price: $1.50 / 1M tokens
- Output price: $7.50 / 1M tokens
- Pay-as-you-go, simple and transparent
-
🌟 Core Features:
- Native Web Search: No tool calling needed, model has built-in web capability
- Real-time Information: Direct access to latest web data
- Stable and Reliable: Reverse-engineered but runs stably
- Plug and Play: No additional Web Search tool configuration needed
-
🛠️ Calling Method:
- Use standard Chat endpoint:
/v1/chat/completions - Compatible with OpenAI API format
- Model automatically handles web requests, no manual tool calling
- Use standard Chat endpoint:
-
⚡ Use Cases:
- Q&A scenarios requiring real-time information
- News and information applications
- Market dynamics analysis
- Fact-checking and verification
- Any tasks requiring latest web data
Grok-4-all and Grok-3-all are models with native web search capabilities. More convenient than traditional tool calling approaches, suitable for applications requiring frequent web information access.
⚡ Grok-4-fast Series Spectacular Launch
September 24, 2025Grok-4-fast - xAI Ultra-long Context Reasoning Model
-
🚀 New Models:
grok-4-fast-reasoning- Reasoning mode, shows complete thought processgrok-4-fast-non-reasoning- Non-reasoning mode, quick responsegrok-4-fast-reasoning-latest- Latest reasoning mode versiongrok-4-fast-latest- Latest standard version
-
💰 Ultra-low Pricing:
- Input price: $0.20 / 1M tokens
- Output price: $0.50 / 1M tokens
- 93%+ price reduction compared to Grok-4 series
- Most cost-effective ultra-long context model ever
-
🧠 Core Features:
- 200K Ultra-long Context: Supports up to 200K tokens context window
- Dual-mode Architecture: Unified model weights, switch reasoning mode via system prompt
- Efficient Reasoning: Saves 40% thinking tokens on average compared to Grok-4
- SOTA Cost Efficiency: Industry-leading cost-effectiveness
-
🌟 Technical Highlights:
- End-to-end tool use reinforcement learning training
- Supports Web Search and X (Twitter) search capabilities
- Built-in code execution capability
- Seamless switching between reasoning and non-reasoning modes
-
🛠️ Calling Method:
- Use standard Chat endpoint:
/v1/chat/completions - Compatible with OpenAI API format
- Pay-as-you-go, simple and transparent
- Use standard Chat endpoint:
Grok-4-fast is xAI’s cost-effective long-context model that maintains Grok-4-level performance while dramatically reducing costs. Perfect for applications requiring large context processing.
💻 Grok Code Fast 1 Code-Specialized Model Launch
September 24, 2025Grok Code Fast 1 - xAI Code Generation and Agent Programming Model
-
🚀 New Model:
grok-code-fast-1 -
💰 Ultimate Cost-effectiveness:
- Input price: $0.20 / 1M tokens
- Output price: $1.50 / 1M tokens
- Cache price: $0.02 / 1M tokens
- Pricing optimized for coding scenarios
-
🧠 Core Features:
- 256K Ultra-long Context: Sufficient for large codebases
- SWE-Bench 70.8%: Outstanding performance on SWE-Bench Verified full set
- High-speed Generation: ~92 tokens/sec throughput
- MoE Architecture: ~314B parameter mixture-of-experts model
-
🌟 Technical Highlights:
- New architecture designed specifically for code tasks
- Multiple programming languages: TypeScript, Python, Java, Rust, C++, Go
- Optimized tool integration: grep, terminal operations, file editing, etc.
- Smart caching: 90%+ cache hit rate in partner workflows
-
🛠️ Recommended Scenarios:
- Building projects from scratch
- Codebase Q&A
- Bug fixing and code refactoring
- Agentic Coding
- IDE integration development
-
⚡ Visible Reasoning Process:
- Response includes reasoning traces
- Developers can guide Grok Code for high-quality workflows
- Enhances code generation controllability and transparency
Grok Code Fast 1 is xAI’s high-performance model built specifically for programming scenarios. Already integrated into mainstream IDEs and programming tools like GitHub Copilot, Cursor, Cline, and Windsurf.
💻 OpenAI Codex Series Models Launch
September 16, 2025GPT-5 Codex - Code Generation Models Optimized for Programming
-
🚀 New Models:
gpt-5-codex-high- High-performance version, benchmarked against GPT-5gpt-5-codex-medium- Medium-performance versiongpt-5-codex-low- Lightweight version
-
💰 Dual Billing Modes:
Pay-as-you-go (suitable for small token conversations):
- High: 10.00/1M output
- Medium: 4.80/1M output
- Low: 2.40/1M output
- High: $0.025/call
- Medium: $0.020/call
- Low: $0.015/call
-
🛠️ Calling Method:
- Use Chat endpoint for calling
- Supports both pay-as-you-go and per-call billing modes
- Flexibly choose billing method based on context length
-
🌟 Core Features:
- Specifically optimized for programming tasks
- Code generation quality comparable to GPT-5
- Supports multiple programming languages and frameworks
- More affordable pricing, extremely cost-effective
Selection Advice: If your context tokens are large, per-call billing is more cost-effective; for short conversations, pay-as-you-go is more economical.
🎨 SeeDream 4.0 API Official Launch
September 11, 2025SeeDream 4.0 - BytePlus Volcano Engine High-Quality Image Generation
-
🚀 New Model:
seedream-4-0-250828 - 🤝 Strategic Partnership: From BytePlus Volcano Engine overseas, APIYI reached official strategic cooperation
-
💰 Pricing Advantage:
- APIYI pricing: 0.03/image)
- Save 35% compared to official
- With 15% top-up bonus: ~¥0.14/image
- Pricing on par with Nano Banana (gemini-2.5-flash-image-preview)
-
🛠️ Calling Method:
- Use standard OpenAI image generation endpoint:
/v1/images/generations - Fully compatible with OpenAI Image API format
- Per-call billing, simple and transparent
- Use standard OpenAI image generation endpoint:
-
🌟 Core Features:
- High-quality image generation: HD 4K output
- Official partnership resources, reliable and stable
- Excellent visual consistency, ~15s per image generation
Detailed usage documentation: SeeDream 4.0 User Guide
📅 August 2025
🎨 Gemini 2.5 Flash Image Generation Model Launch
August 27, 2025Gemini 2.5 Flash Image Preview - Google’s Strongest Image Generation Model
-
🚀 New Model:
gemini-2.5-flash-image-preview - 🏷️ Model Code Name: Nano Banana model
-
💰 Pricing Advantage:
- APIYI pricing: 0.04/image)
- With top-up bonus + exchange rate benefits: ~¥0.14/image
- Save about 50% compared to official (50% off official price)
-
💸 Competitive Comparison:
- Better advantage than gpt-image-1
- Pricing on par with flux-kontext-pro
- Higher than reverse-engineered sora_image
-
⚡ Performance Advantages:
- Fast generation speed, average only 10 seconds
- Faster response time than OpenAI series
- Google’s latest and strongest image generation/editing technology
-
🛠️ Calling Method:
- Use chat completion endpoint, compatible with gpt-4o-image
- Simply replace model name to use
- Perfectly compatible with sora_image calling method
Detailed usage documentation: Gemini Image Generation User Guide
🧠 DeepSeek V3.1 Mixed Reasoning Mode Launch
August 26, 2025DeepSeek V3.1-250821 - Revolutionary Mixed Reasoning Mode
-
🚀 New Model:
deepseek-v3-1-250821 -
💰 Pricing Advantage:
- Input price: 0.56)
- Output price: 1.68)
- Save about 11% compared to official pricing
- Supports dual-mode calling with flexible billing
-
🧠 Core Features:
- Think Mode: Deep reasoning, shows complete thought process
- Non-Think Mode: Quick response for everyday tasks
- 128K ultra-long context window
- Supports Anthropic API format compatibility
-
🌟 Technical Highlights:
- 840B token continuous pre-training optimization
- New tokenizer with improved encoding efficiency
- Beta Function Calling tool support
- Enhanced agent and tool usage capabilities
-
⚡ Mode Switching:
deepseek-chat- Non-Think quick modedeepseek-reasoner- Think reasoning mode- Official supports “DeepThink” one-click switching
DeepSeek V3.1 is the first large model supporting mixed reasoning mode, allowing users to flexibly choose reasoning depth based on task complexity, balancing efficiency and quality.
🤖 Kimi K2 Official Release Integration
August 25, 2025Kimi K2-250711 - Official Volcano Engine Partnership Version
-
🚀 New Model:
kimi-k2-250711 -
🔄 Replaces Model: Replaces
kimi-k2-0711-previewpreview version -
🤝 Official Partnership:
- Official Volcano Engine authorized partnership version
- Direct connection to official interface, significantly improved stability
- Goodbye to preview version instability
-
💰 Pricing Advantage:
- Continues 15% lower than official pricing
- With top-up bonuses, equivalent to 30% off official
- Official release performance at preview pricing
-
🌟 Core Features:
- Excellent Chinese understanding and generation
- Strong long-text processing capability
- More stable and reliable API response
- Fully compatible with original calling method
Kimi K2-250711 is the official release version in partnership with Volcano Engine. Compared to preview version, it has significant improvements in stability and performance. Users are recommended to switch promptly.
🚀 GPT-5 Full Series Spectacular Launch
August 8, 2025GPT-5 Full Series Official Release - OpenAI’s Strongest Model
-
🎯 Full Series Models:
gpt-5- Flagship versiongpt-5-2025-08-07- Specific versiongpt-5-chat-latest- Latest chat versiongpt-5-mini/gpt-5-mini-2025-08-07- Lightweight versiongpt-5-nano/gpt-5-nano-2025-08-07- Ultra-lightweight version
-
💰 Pricing Advantage:
- Completely aligned with OpenAI official pricing
- Top-up bonus + exchange rate benefits, about 20% off official
- Input price: 0.05 / 1M tokens
- Output price: 0.40 / 1M tokens
-
🛠️ Technical Features:
- Supports official
/v1/responsesendpoint calling - Temperature parameter
temperaturemust be set to 1 or omitted (official limitation) - Use
max_completion_tokensinstead ofmax_tokens gpt-5-chat-latestcalled via/v1/chat/completions
- Supports official
-
⚡ Performance Highlights:
- Comprehensively exceeds GPT-4 series reasoning capabilities
- Stronger context understanding and long-text processing
- Significantly improved code generation quality
- Multilingual capabilities reach new heights
Important Note: GPT-5 series models have specific parameter requirements. Please configure according to the above technical features to ensure normal calling.
🎨 Claude Opus 4.1 Performance Upgrade Launch
August 7, 2025Claude Opus 4.1 - Code Capability Breakthrough
-
🚀 New Model:
claude-opus-4-1-20250805 -
💰 More for Same Price:
- Pricing completely aligned with Anthropic official
- Better cost-effectiveness with top-up bonuses
- Same price, enjoy stronger performance
-
🌟 Core Upgrades:
- SWE-bench Verified reaches 74.5%, significantly improved code capability
- Compared to Opus 4, greatly enhanced reasoning and code generation quality
- Maintains 200K ultra-long context window
- Further optimized multilingual understanding and generation
-
⚡ Performance Highlights:
- Industry-leading code debugging and fixing capabilities
- 15%+ improvement in complex reasoning task accuracy
- API response speed maintains industry-leading level
- Perfectly compatible with all Claude series features
Claude Opus 4.1 is an iterative optimization version based on Opus 4. While maintaining original powerful capabilities, it has undergone specialized optimization for code generation and reasoning tasks.
🎯 OpenAI Open Source Models Official Launch
August 6, 2025OpenAI’s First Open Source Large Models Spectacular Release
-
🚀 New Models:
gpt-oss-120b- 117B parameters, performance comparable to o4-minigpt-oss-20b- 21B parameters, can run on edge devices
-
💰 Pricing Advantage:
- Significantly lower pricing than DeepSeek R1 and V3
- Extremely cost-effective with top-up bonuses
- Significantly reduced enterprise usage costs
-
🌟 Core Features:
- 128K ultra-long context window
- Supports low/medium/high three-level reasoning modes
- MoE architecture, extremely efficient inference
- Apache 2.0 open source license
-
⚡ Performance Highlights:
- gpt-oss-120b runs on single 80GB GPU
- gpt-oss-20b supports 16GB memory device deployment
- Outstanding performance in programming, math, tool calling tasks
This is OpenAI’s first open source model release since GPT-2, marking an important milestone in the AI open ecosystem.
📅 July 2025
⚠️ GPT-4.5-Preview Model Discontinued
July 14, 2025GPT-4.5-Preview Officially Stops Service
- 📌 Discontinued model:
gpt-4.5-preview - 🔄 Recommended replacement:
gpt-4.1 - 📅 Effective date: July 14, 2025
- 📖 Official announcement: OpenAI Deprecations
Developers please update code promptly, replace
gpt-4.5-preview with gpt-4.1 to avoid service interruption.🚀 Kimi K2 Model Launch
July 15, 2025Kimi K2 Preview Version Official Launch
- 📌 Model name:
kimi-k2-0711-preview - 💰 Pricing advantage: 15% lower than official
- 🎁 With top-up bonus, equivalent to 30% off official
- 🌟 Features: Better Chinese understanding and generation capabilities
🤖 Grok-4 Model Launch
July 12, 2025xAI Grok-4 Official Integration
- 📌 Calling name:
grok-4 - 🧠 Special feature: Built-in reasoning mode
- 📖 Recommended to check official docs for detailed usage
💸 Claude Full Series 20% Price Reduction
July 8, 2025Claude Models Major Discount
- 🎉 Full series 20% price reduction
- ✅ No speed limit, no account ban
- 💱 Exchange rate advantage + top-up bonus
- 🚀 Providing more economical AI services for users
📅 June 2025
🌟 Gemini 2.5 Official Release
June 18, 2025Gemini 2.5 Series Completely Upgraded
- 📌 Recommended models:
gemini-2.5-pro/gemini-2.5-flash - ⚡ More stable performance, recommended to replace preview or exp versions
- 🎯 2M ultra-long context support
🎊 O3 Model Major 80% Price Reduction
June 13, 2025O3 Model Biggest Price Drop Ever
- 💥 Price reduction: 80%
- 💰 New pricing:
- Input: $3/M Tokens
- Output: $12/M Tokens
- 🆕 New
o3-promodel added - 📍 Only available via
/v1/responsesendpoint
🖼️ GPT-Image-1 Price Reduction
June 7, 2025More Affordable Image Generation
- 📌 Model:
gpt-image-1 - 💸 Significant price reduction, more competitive image generation service
- 🎨 Maintain high-quality output while reducing costs
🔔 Subscribe to Updates
Join Community
Join Telegram community for real-time update notifications
View Pricing
Check latest pricing information for all models
📊 Update Statistics
This Month Highlights (October 2025)
- ✨ Version Update: Gemini 2.5 Flash latest dated version preview-09-2025 launched, Google’s latest optimization
- 🍌 Image Upgrade: Gemini 2.5 Flash Image official release, 10 aspect ratios, custom resolution support
- 🎬 Video Revolution: Sora 2 video generation launched, audio-video sync, watermark-free output, from $0.15/call
- 💻 Intelligence Upgrade: GLM-4.6 strong release, 200K context, comprehensive code and reasoning improvements
Last Month Highlights (September 2025)
- 🚀 World’s Strongest: Claude Sonnet 4.5 spectacular launch, SWE-bench 77.2%, world’s best coding model
- 🧠 Quick Response: DeepSeek-V3.2-exp early launch, first to integrate latest experimental version
- 🤖 Domestic Power: GLM-4.5 series fully launched, priced below official
- 🌐 Web Capability: Grok-4-all / Grok-3-all launched, native web search, no tool calling needed
- ⚡ Ultra-long Context: Grok-4-fast series launched, 200K tokens context, 93%+ price reduction
- 💻 Code-Specialized: Grok Code Fast 1 released, SWE-Bench 70.8%, integrated into mainstream IDEs
- 💻 Programming Optimization: OpenAI Codex series launched, optimized for programming scenarios, dual billing modes
- 🎨 Image Generation: SeeDream 4.0 officially launched, BytePlus Volcano Engine strategic partnership, 35% off official
Historical Milestones (August 2025)
- 🎨 Image Generation: Gemini 2.5 Flash Image Preview launched, Google’s strongest image model, 30% off official
- 🧠 Innovation Breakthrough: DeepSeek V3.1 mixed reasoning mode, first large model supporting Think/Non-Think dual modes
- 🤖 Official Partnership: Kimi K2-250711 official release, official Volcano Engine authorization, significantly improved stability
- 🚀 New Models: GPT-5 full series officially launched, 20% off with official pricing
- 🎨 Performance Upgrade: Claude Opus 4.1 performance upgrade, more for same price
- 🎯 Major Release: OpenAI’s first open source large models (gpt-oss-120b, gpt-oss-20b)
- 💸 Pricing Advantage: New models priced lower than DeepSeek R1/V3
- 🚀 Performance Breakthrough: Open source models reach commercial-grade performance
- 🌐 Ecosystem Expansion: Support local deployment and edge computing
Historical Milestones
- October 2025: Gemini 2.5 Flash latest dated version preview-09-2025 launched; Gemini 2.5 Flash Image official release; Sora 2 video generation launched, OpenAI’s revolutionary video model; GLM-4.6 strong release, Zhipu AI code and reasoning enhanced edition
- September 2025: Claude Sonnet 4.5 spectacular launch, world’s strongest coding model; DeepSeek-V3.2-exp early launch; GLM-4.5 series fully launched; Grok web search models launched, native web capability; xAI Grok-4-fast series launched, 200K ultra-long context record low price; Grok Code Fast 1 code-specialized model released; OpenAI Codex series released, optimized for programming; SeeDream 4.0 officially launched, BytePlus Volcano Engine strategic partnership reached
- August 2025: Gemini 2.5 Flash Image Preview Google’s strongest image model launched; DeepSeek V3.1 mixed reasoning mode launched; Kimi K2 official release integrated; GPT-5 full series spectacular launch; First integration of OpenAI open source models; Claude Opus 4.1 performance upgrade released
- July 2025: Breakthrough 200+ model support
- June 2025: O3 model biggest price reduction ever
- May 2025: Support Gemini 2.5 series
Update Frequency: This page is regularly updated. Recommend bookmarking and checking frequently. All price adjustments and model updates are announced here first.
Pro Tip: New model launches typically have special offers initially. Recommend trying them out promptly. We continuously optimize pricing strategy based on market feedback.