Changelog
July 23, 2025
- Added support for Qwen Code, leveraging all large language models on the Aihubmix platform.
July 4, 2025
- Added llms.txt support — a single link to help your LLM assistant quickly navigate the entire model ecosystem.
June 29, 2025
- Added Gemini Cli documentation, supporting Gemini model usage in your terminal.
- The OpenAI Responses API now supports Code Interpreter and Remote MCP calls.
June 26, 2025
- Added Universal Image Generation API, supporting multiple mainstream models including OpenAI, Ideogram, Stability, and Google Imagen.
June 18, 2025
- Added HTTP Status Codes documentation to help users better understand error information.
June 15, 2025
- Added Veo 3.0 reverse access method, with a total cost of just $0.41 per video generation.
June 12, 2025
- Integrated Claude, enabling stable use in China.
June 9, 2025
- Supports displaying OpenAI Reasoning summaries, applicable to the responses API.
June 5, 2025
- Added Gemini implicit caching feature, supporting automatic hit and hit feedback. Developers can determine hit status through usage_metadata. Cost savings are not guaranteed, and actual effects depend on request structure and usage scenarios.
May 31, 2025
Comprehensive Claude 4 New Features Support
- ⏳ New Cache Duration: Claude 1-hour cache durationbeta support.
- 🎉 New Text Editor Tool: Claude 4 models support the new
text_editor_20250429tool type andstr_replace_based_edit_tooltool name - 🚫 Refusal Stop Reason: Introduced new
refusalstop reason for content the model declines to generate for safety reasons - 🧠 Extended Thinking: Claude 4 models return a summary of the full thinking process, providing complete intelligence benefits of extended thinking
- 🔄 Interleaved Thinking: Support for interleaving tool use with extended thinking, enabling more natural conversation experiences (Beta feature)
- ⚠️ Deprecated Features:
- Claude 4 text editor tool no longer supports
undo_editcommand - Removed
token-efficient-tools-2025-02-19support (Claude 3.7 only) - Removed
output-128k-2025-02-19support (Claude 3.7 only)
- Claude 4 text editor tool no longer supports
- 📚 Updated comprehensive migration guide and code examples to help users smoothly migrate from Claude 3.7 to Claude 4
May 22, 2025
- Added support for Dify plugin, enabling seamless integration of Aihubmix models in Dify.
May 17, 2025
- The codex-mini-latest model, designed specifically for programming tasks, is now supported and can be accessed via the Responses API endpoint or the Codex CLI.
- Support Google Imagen 3.0 for image generation and Veo 2.0 for video creation, enriching your creative formats.
- gemini-2.0-flash-exp has been upgraded to gemini-2.0-flash-preview-image-generation
May 9, 2025
Added Ideogram AI V3 interface, the most advanced image generation model.
May 6, 2025
Added AiHubMix CLI, supporting command-line management of keys, account information, and available model lists.
April 26, 2025
- The long-awaited OpenAI image generation interface is officially launched, supporting text-to-image generation and image-to-image editing.
- Supports native invocation for Gemini models
April 24, 2025
Integrated with 3 core APIs from Jina AI, empowering you to effortlessly build powerful intelligent agents. The three APIs are: Vector Embeddings, Rerank, and Deep Search.
April 22, 2025
GPT 4o Image Reverse Interface Early Access.
April 20, 2025
OpenAI Responses API is supported, offering more comprehensive tool integration.
April 17, 2025
Added OpenAI CodeX Cli support, Tell Codex CLI what to build, fix, or explain, then watch it bring your ideas to life.
April 12, 2025
Added
:surfing to the model id to enable search functionality for any model BetaApril 9, 2025
Support Claude prompt caching, high-frequency prompts can save up to 76% of costs!
April 7, 2025
Added support for Ideogram AI drawing interface, featuring powerful text-to-image capabilities, including image generation, remix, partial editing and upscaling.
April 5, 2025
New docs published, bring you a better experience
March 30, 2025
Added support for Claude Text Edit Tool
March 24, 2025
Enabled new Trident Logo
March 16, 2025
Added native search functionality support for OpenAI and Google Gemini models; future plans include extending this interface to support third-party search
March 15, 2025
Added new models: gpt-4o-mini-search-preview and gpt-4o-search-preview
March 7, 2025
o1 and o3-mini prices reduced by 10%, now aligned with official pricing
March 6, 2025
aihubmix-DeepSeek-R1 price increased 7x due to Microsoft price increase; recommend using Volcano’s DeepSeek-R1 for stability and affordability. Added new models: qwen-qwq-32b and qwen2.5-vl-72b-instruct
February 28, 2025
Claude models price reduced by 15%; Added new model gpt-4.5-preview (Note: extremely expensive, please use with caution)
February 26, 2025
Improved Deepseek stability; Most stable from ByteDance, recommended models: DeepSeek-R1 and DeepSeek-V3
February 25, 2025
Added model claude-3-7-sonnet-20250219
February 24, 2025
- gpt-4o occasionally experiencing slow responses due to vendor issues; temporarily recommend using gpt-4o-2024-11-20
- Perplexity API temporarily offline due to complex official pricing structure exceeding our platform’s pricing structure; will relaunch after price adjustment
- ByteDance official limited-time discount ended, DeepSeek-R1 price has increased
- Added model details page and parameter information
February 23, 2025
- ByteDance official limited-time discount ended, DeepSeek-V3 price has increased; expecting R1 price to increase soon, we will adjust accordingly
February 18, 2025
- Added new model: kimi-latest (Official pricing based on input length tiers 8k/32k/128k, our platform uses 32k tier as standard, price-sensitive users please note)
- Optimized website structure
- Merged logs page into usage statistics page
- Moved announcements to model marketplace page
- Moved settings under profile picture
- aihubmix-DeepSeek-R1 price reduced by 50%
- Added new models: gemini-2.0-pro-exp-02-05-search, gemini-2.0-flash-exp-search with integrated Google official search functionality
- Added new models: gemini-2.0-flash, gemini-2.0-pro-exp-02-05, gemini-2.0-flash-lite-preview-02-05
- Added new models: o3-mini, o1 (Note: backend billing ~10% higher than official due to limited accounts)
February 4, 2025
- o1 model: OpenAI official does not support stream parameter
- o3-mini does not support temperature parameter; added new parameter “Reasoning effort” accepting “low, medium, high” (defaults to medium)
February 1, 2025
Feature: Added support for OpenAI voice model input/output functionality. Available on api.aihubmix.com server, main server support coming after 1 week of stability. Overall backend billing matches official rates. Currently, logs only show text tokens; voice costs not yet displayed but functionality unaffected.New models added:
- o3-mini, o1 (Note: backend billing ~10% higher than official due to limited accounts)
- aihubmix-DeepSeek-R1 (Recommended, more stable)
- qwen-max-0125 (Qwen2.5-Max), sonar-reasoning
- deepseek-ai/DeepSeek-R1-Zero and deepseek-ai/DeepSeek-R1, deepseek-r1-distill-llama-70b
- aihub-Phi-4
- Doubao-1.5-pro-256k, Doubao-1.5-pro-32k, Doubao-1.5-lite-32k, Doubao-1.5-vision-pro-32k
- sonar, sonar-pro (latest from Perplexity AI)
- gemini-2.0-flash-thinking-exp-01-21
- deepseek-reasoner (aka DeepSeek-R1)
- MiniMax-Text-01
- codestral-latest (Mistral’s new code model - Codestral 25.01)
January 23, 2025
New models added:
- aihub-Phi-4
- Doubao-1.5-pro-256k, Doubao-1.5-pro-32k, Doubao-1.5-lite-32k, Doubao-1.5-vision-pro-32k
- sonar, sonar-pro (latest from Perplexity AI)
- gemini-2.0-flash-thinking-exp-01-21
- deepseek-reasoner (aka DeepSeek-R1)
January 19, 2025
- Added Perplexity AI API models; currently only available on preview server api.aihubmix.com, will update to main server aihubmi.com if stable
- api.aihubmix.com is preview server, new features update here first, typically moving to main server aihubmix.com after 1 week of stability
- MiniMax-Text-01
- codestral-latest (Mistral’s new code model - Codestral 25.01)
- gpt-4o-zh: automatically translates any language input to English for the model, model output automatically translated back to Chinese; feature in testing phase, limited concurrent requests, only supports gpt-4o model
January 6, 2025
- Added gemini-2.0-flash-exp-search with Google native web search functionality; official gemini 2.0 flash model’s web search requires additional parameters, aihubmix integrated this - just add ‘search’ to model name to use
- Added new model deepseek-ai/DeepSeek-V3
January 1, 2025
- Added model marketplace page replacing original model/pricing page
December 30, 2024
- Fixed gemini-2.0-flash-thinking-exp-1219 model only outputting thoughts without answers
- Fixed issue with balance reminder emails not being received
December 22, 2024
- Added usage statistics page and recharge record page
- Added Doubao series models: Doubao-lite-128k, Doubao-lite-32k, Doubao-lite-4k, Doubao-pro-128k, Doubao-pro-256k, Doubao-pro-32k, Doubao-pro-4k
- Added new models: gemini-2.0-flash-thinking-exp-1219, gemini-2.0-flash-exp, aihubmix-Mistral-Large-2411, aihubmix-Llama-3-3-70B-Instruct, grok-2-1212, grok-2-vision-1212
- Added new models: gemini-exp-1206, llama-3.3-70b-versatile, learnlm-1.5-pro-experimental
December 14, 2024
- Added new models: gemini-2.0-flash-exp, aihubmix-Mistral-Large-2411, aihubmix-Llama-3-3-70B-Instruct
December 8, 2024
- Added new models: gemini-exp-1206, llama-3.3-70b-versatile, learnlm-1.5-pro-experimental
- Added usage statistics page
November 21, 2024
- Recently added models: gpt-4o-2024-11-20, step-2-16k, grok-vision-beta
- Qwen 2.5 Turbo Million Context Model: qwen-turbo-2024-11-01
November 7, 2024
- Compatible with Claude native SDK, v1/messages interface now supported
- Cache and computer usage functions for Claude native interface not yet supported (prompt caching and computer use); will continue improving these features in next two weeks
November 5, 2024
- Added new model: claude-3-5-haiku-20241022
- Added Elon Musk’s x.ai latest model grok-beta
October 23, 2024
- Added new model: claude-3-5-sonnet-20241022
October 10, 2024
OpenAI’s latest caching feature now online. Currently supports:
- GPT-4o
- GPT-4o-mini
- o1-preview
- o1-mini
October 3, 2024
- Reduced backend billing for gpt-4o model, synchronized with official pricing
- Added new models: aihubmix-Llama-3-2-90B-Vision, aihubmix-Llama-3-70B-Instruct
- Added Cohere’s latest models: aihubmix-command-r-08-2024, aihubmix-command-r-plus-08-2024
September 19, 2024
- Added new models: whisper-large-v3 and distil-whisper-large-v3-en
- Note: Whisper model billing based on input seconds; current page pricing display issue to be fixed. Backend billing accurate, fully synchronized with OpenAI official charges
September 13, 2024
- Added models o1-mini and o1-preview Note: These latest models require input parameter changes. Some wrapper software may report errors if default parameters not updated.
- system field: 400 error
- tools field: 400 error
- image input: 400 error
- json_object output: 500 error
- structured output: 400 error
- logprobs output: 403 error
- stream output: 400 error
- o1 series: 20 RPM, 150,000,000 TPM, very low, 429 errors may occur anytime
- Others: temperature, top_p, and n fixed at 1; presence_penalty and frequency_penalty fixed at 0
September 10, 2024
- Added new model: mattshumer/Reflection-Llama-3.1-70B (reportedly strongest fine-tuned version of llama3.1-70b)
- Increased Claude-3 model pricing to maintain stable supply, currently 10% more expensive than direct official API use; prices will gradually decrease
- Enhanced concurrent capabilities of OpenAI series models, theoretically supporting unlimited concurrency
August 11, 2024
- Added new models: Phi3medium128k, ahm-Phi-3-medium-4k, ahm-Phi-3-small-128k
- Improved stability of Llama-related models
- Further optimized compatibility of Claude models
August 7, 2024
- Added OpenAI’s newly updated 4o version gpt-4o-2024-08-06, see https://platform.openai.com/docs/guides/structured-outputs
- Added Google’s latest model: gemini-1.5-pro-exp-0801
August 4, 2024
- Added online direct payment recharge
- Fixed Claude multi-turn conversation format error: messages roles must alternate between “user” and “assistant”, but found multiple “user” roles in a row
- Optimized index issue when using function features of Claude models
- Backup server https://orisound.cn will be fully decommissioned on September 7; please switch to main server https://aihubmix.com or backup server https://api.aihubmix.com
July 27, 2024
- Added support for Mistral Large 2, model name: Mistral-large-2407 or aihubmix-Mistral-large-2407
- System optimization
July 24, 2024
- Added latest llama-3.1 models: llama-3.1-405b-instruct, llama-3.1-70b-versatile, and llama-3.1-8b-instant
July 20, 2024
- Fixed pricing calculation issue for gpt-4o-mini model: Text input price: Official gpt-4o-mini model text input price is 1/33 of gpt-4o model price Image input price: Official gpt-4o-mini model image input price equals gpt-4o model price
- For accurate price calculation, we multiply token count for gpt-4o-mini model image input by 33x to align with official price
-
More details at OpenAI Official Pricing
July 19, 2024
- Added support for gpt-4o-mini model, backend billing synchronized with official
July 15, 2024
- Supports official API parameter include_usage, passing parameters can return usage in stream mode, see Official Documentation
July 14, 2024
- New nextweb version supports calling non-OpenAI models Call Non-OpenAI Models on Our Site
- Added backend billing for Alibaba Qwen models, overall cost ~10% higher than Alibaba Cloud official
- Optimized Azure OpenAI output compatibility with OpenAI interface
- Supports Claude-3’s tool calling
- Added many new models, see Settings for available models
July 3, 2024
- Overall backend interface optimized
- Each log request record now shows model unit price at time of request
- Added model and pricing page Model/Pricing
June 20, 2024
- Latest claude-3-5-sonnet-20240620 now supported, see usage instructions at How to Call Non-OpenAI Models on Our Site
June 18, 2024
- Backend log page now supports downloading usage request records
June 16, 2024
- Reduced probability of randomly hitting Azure OpenAI, now almost negligible
June 13, 2024
- Lowered fees for Claude-3 related models (Claude 3 Haiku, Claude 3 Sonnet, Claude 3 Opus) to align with official backend charges; current retail price ~14% off official price
June 10, 2024
- Overall service architecture upgrade, all servers and data migrated to Microsoft Azure
- Future plans include secondary deep development and optimization based on open-source version of one API (commercial license obtained through sponsorship)
- Log data too large (over 100 million request logs) for temporary migration; contact customer service for previous log queries
- Optimized token billing for gpt-4o, tokenizer changed from cI100k_base to 0200k_base; results in decreased token count for streaming requests in Chinese, Korean, and Japanese
June 8, 2024
- Added Alibaba’s latest open-source models:
- alibaba/Qwen2-7B-Instruct
- alibaba/Qwen2-57B-A14B-Instruct
- alibaba/Qwen2-72B-Instruct
May 20, 2024
- Added new model gemini-1.5-flash
- Added new model gpt-4o
- Error accessing recharge page in Jiangsu region due to telecom hijacking of recharge domain, please contact customer service for recharge
- Added llama3 (llama3-70b-8192, llama3-8b-8192) gemini-1.5-pro, command-r, command-r-plus
- Claude-3 model supply restored; currently connecting to Claude-3 endpoints deployed on AWS and Google Cloud
- Claude-3 model and price backend billing 10% more expensive than official to cover server costs and team expenses
- Prices will gradually decrease to ~5% or lower as call volume increases
- Current concurrency needs testing; will apply for higher concurrent calls as usage increases