Skip to main content
Welcome to MixRoute’s changelog page. Here you can find important information about our latest model launches, pricing adjustments, feature updates, and more.
Stay updated: press Ctrl+D to bookmark this page and get the latest model launches and promotional information as soon as possible.

​
🔥 Latest Updates

Gemini 4 Argon

Reasoning for complex coding, mathematics, and agent workflows, with a shared 1M-token context window.

GPT 6.1 Sol

Text and image reasoning for coding and professional work. Function calling uses the Responses API.

Claude Opus 5.5

Long-running coding agents and knowledge work, with always-on adaptive thinking and up to 128K output.

Claude Haiku 5.5

Low-latency classification, extraction, routing, and coding subagents, with adaptive thinking.

Grok 4.7

Text and image input, configurable reasoning, and tool workflows through the Responses API.

DeepSeek V4.1 Flash

Native visual understanding for coding, agents, and multimodal analysis.

​
🔧 Update Announcements

​
2026-10-15 Model catalog and documentation sync

This update standardizes the displayed model count at 250 and adds changelog entries for models already supported by MixRoute. The date is the documentation update date, not a first-availability or vendor release date.

​
Integration notes

  • Gemini 4 Argon shares one 1,048,576-token window across input, thinking, and output; long decode is a separate workflow, not a 1M-token single Chat Completions response.
  • Use the Responses API for GPT 6.1 Sol function calling. Its lowest reasoning effort is low; none and minimal are not supported.
  • Claude Opus 5.5 always uses adaptive thinking. Use output_config.effort; forced tool choices any and tool are not supported.
See Popular Models for model selection and the Model Marketplace for current availability and pricing.

​
2026-08-26 GLM-5.3 Flash Release

Z.ai released glm-5.3-flash, its first natively multimodal GLM-5 model:
  • 1M-token context for long coding, document, and agent workflows
  • Text, image, video, and file understanding
  • Efficient 320B-parameter MoE architecture with 18B active parameters
  • Hybrid sparse and linear attention for lower long-context serving cost
Source: Z.ai GLM-5.3 Flash announcement

​
2026-08-14 GLM-5.3 Release

Z.ai released glm-5.3, a flagship always-thinking model built for complex coding and long-horizon agents:
  • 1M-token context and up to 128K output
  • low, high, and max reasoning effort; max is recommended for coding
  • Stronger coding, autonomous execution, and security analysis than GLM-5.2
  • OpenAI-compatible Chat Completions through MixRoute
Source: Z.ai GLM-5.3 announcement

​
2026-08-13 DeepSeek V4 Pro General Availability

DeepSeek released the general-availability version of deepseek-v4-pro:
  • Stronger production agent and coding performance
  • Native OpenAI Responses API support
  • low, high, and max reasoning effort
  • 1M-token context and up to 384K output
Source: DeepSeek V4 Pro GA announcement

​
2026-08-12 Grok 4.6 Release

xAI released grok-4.6 for coding, agentic tasks, and knowledge work:
  • 500K context window
  • Text and image input with text output
  • Function calling and structured output
  • low, medium, high, and xhigh reasoning effort
Source: xAI release notes

​
2026-07-23 Moonshot Kimi K3 & New Gemini Series Launch

  • Moonshot: New flagship model kimi-k3 is now available, replacing kimi-k2.6 as the new Kimi series flagship, with upgraded reasoning and multimodal capabilities
  • Google: Added gemini-3.6-flash, gemini-3.5-flash-lite and other new models

​
2026-07-10 GPT 5.6 New Flagship Series Launch

OpenAI’s new GPT-5.6 series is now available, introducing a new naming scheme with major breakthroughs in reasoning efficiency, frontend design, and tool calling:
  • OpenAI: New GPT-5.6 series models launched with a new naming scheme:
    • gpt-5.6-sol — Flagship reasoning model, supports Pro Mode deep reasoning (reasoning.mode: "pro")
    • gpt-5.6-terra — Cost-effective flagship, balancing performance and cost
    • gpt-5.6-luna — High-efficiency, high-volume, ideal for production workloads
    • New reasoning.effort: max highest reasoning level
    • Supports Programmatic Tool Calling, Multi-agent (beta), explicit Prompt Caching, persisted reasoning (reasoning.context)
    • Token efficiency significantly improved, frontier quality with fewer tokens

​
2026-07-01 New Upgrade to the Video and Image Model Library

We have fully introduced ByteDance’s latest video and image generation models, giving you a broader range of visual creation capabilities:
  • New Volcano Engine / ByteDance models launched:
    • Newly launched flagship SeeDream 5.0 (seedream-5-0-260128) and its lightweight version SeeDream 5.0 Lite (seedream-5-0-lite-260128).
    • Added stable and practical SeeDream 4.x-level models, including seedream-4-0-250828 (version 4.0) and seedream-4-5-251128 (version 4.5).
    • Added the lightweight flagship video model Seedance 2.0 Mini (dreamina-seedance-2-0-mini-260615).
    • All newly added models are fully integrated and support high-quality output, making it easy for multilingual developers to call them flexibly across different application scenarios.
  • Anthropic: Also launched the flagship model Claude Sonnet 5 (claude-sonnet-5), with a 1M-token ultra-long context, suitable for coding, agents, and enterprise workflows.

​
2026-06-05 Major Upgrade to the Video Model Library

Based on the latest progress in the platform’s video generation capabilities, we have fully synchronized and updated the Seedance model series:
  • ByteDance (Dreamina/Seedance):
    • Newly launched the latest flagship dreamina-seedance-2-0-260128 (Seedance 2.0) and dreamina-seedance-2-0-fast-260128 (fast version).
    • Added support for the high-quality professional version seedance-1-5-pro-251215 (1.5 Pro).
    • Aligned and updated the 1.0 series, including seedance-1-0-pro-250528 and seedance-1-0-pro-fast-251015.
    • Also upgraded the developer documentation for video task submission and querying, and added English and Traditional Chinese code examples and comparison tables to make integration easier for multilingual developers.

​
2026-05-27 Full Model Library Synchronization Update

Based on the latest platform API status, we have fully synchronized and updated the models supported by the platform:
  • OpenAI: Launched the gpt-5.5 flagship series, the gpt-5.4 series, the reasoning model o4-mini, and the image model gpt-image-2
  • Anthropic: Launched Anthropic model updates, including claude-opus-4-7, claude-sonnet-5, and claude-haiku-4-5-20251001
  • Google: Launched the Gemini 3.1 and 3.5 series, including gemini-3.5-flash, gemini-3.1-pro-preview, and more
  • DeepSeek: Launched the V4 series, including deepseek-v4-pro and deepseek-v4-flash
  • xAI: Launched the Grok 4 series, including grok-4.20-beta-0309-reasoning, and more
  • Chinese models: Launched key flagship models such as qwen3.7-max, glm-4-plus, and kimi-k2.6
Update frequency: this page is updated regularly. We recommend bookmarking it and checking back often. All pricing adjustments and model updates will be announced here as soon as possible.
Newly launched models often come with special promotions during the initial launch period, so we recommend trying them early.