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

# Axon Changelog

> Product updates and announcements for MatterAI Axon LLM

<Update label="December 16th, 2025">
  ## Axon Model 2.0

  Major performance upgrade with reduced pricing and state-of-the-art benchmarks.

  * **2x Faster Speed**: Optimized inference latency for rapid responses.
  * **89% LiveCodeBench**: Achieving top-tier accuracy in code generation benchmarks.
  * **New Pricing**: Reduced to $1/1M input (was $2) and $4/1M output (was $6).

  ## Core Improvements

  * **Parallel Processing**: Enabled parallel client functionality for improved concurrent request handling.
  * **Rate Limiting**: Refined strategies and fixed context propagation in API responses.
</Update>

<Update label="November 20th, 2025">
  ## Cortex Intelligence Upgrade

  Updated Cortex modules with advanced planning capabilities.

  * **Enhanced Reasoning**: New planning prompts and task memory helper.
  * **Better Execution**: Significantly improved agent reliability in complex tasks.
</Update>

<Update label="October 10th, 2025">
  ## Axon Playground Axon Playground is now available to test and explore the

  capabilities of Axon models in the dashboard.
</Update>

<Update label="October 5th, 2025">
  ## New Axon-Mini model released Axon Mini 1 is now available for everyday

  low-effort tasks with deep reasoning. Read more [here](/axon-mini).
</Update>

<Update label="September 30th, 2025">
  ## Deep Reasoner v2.0 Deep Reasoner v2.0 is now available with massive boost

  in speed, reasoning and accuracy.
</Update>

<Update label="September 23rd, 2025">
  ## Global Rate Limits Global API Rate Limits are not applicable based on your

  Tier. More details [here](/axon-ai/rate-limits).
</Update>

<Update label="September 11th, 2025">
  ## Tool call usage response format fixed Tool call usage response format is

  now fixed for AI editors and tools to consume.
</Update>

<Update label="September 7th, 2025">
  ## Fix reasoning summary in response Breaking reasoning summary in response

  format is not fixed.
</Update>

<Update label="September 3rd, 2025">
  ## Support for `v1/messages` API for Claude Code Deep Reasoner now outputs

  less tokens to reduce the cost of the model.
</Update>

<Update label="September 1st, 2025">
  ## Reduced tokens for Deep Reasoner Deep Reasoner now outputs less tokens to

  reduce the cost of the model.
</Update>

<Update label="August 25th, 2025">
  ## Improved Response Speed MatterAI Axon now supports improved response speed

  to provide a high level overview of the reasoning process. This allows you to
  balance between accuracy and performance.
</Update>

<Update label="August 20th, 2025">
  ## Web Search & Web Fetch Tool Calling

  MatterAI Axon now supports web search and web fetch tool calling to provide a high level overview of the reasoning process. This allows you to balance between accuracy and performance.

  * `web_search` - LLM will use web search to find relevant information
  * `web_fetch` - LLM will scrape the provided URL to find relevant information

  <Note>
    Tool Calling is always available and enabled by default.
  </Note>
</Update>

<Update label="August 10th, 2025">
  ## Reasoning Summary in Response

  MatterAI Axon now supports reasoning summary in response to provide a high level overview of the reasoning process. This allows you to balance between accuracy and performance.

  * `auto` - AI will decide whether to provide a summary or not
  * `none` - AI will never provide a summary

  ```json theme={null}
  {
    "reasoning": {
      "summary": "auto"
    }
  }
  ```

  <Note>
    Reasoning summary is always enclosed in a `<reasoning_start>` REASONING `<reasoning_end>` tag.
  </Note>
</Update>

<Update label="August 5, 2025">
  ## Reasoning Effort Options

  MatterAI Axon now supports reasoning effort options to control the level of reasoning and analysis performed by the AI model. This allows you to balance between accuracy and performance.

  * `None` - No reasoning, fastest
  * `Low` - Fastest, least accurate
  * `Medium` - Good balance
  * `High` - Most accurate, slowest

  ```json theme={null}
  {
    "reasoning": {
      "effort": "low | medium | high | none"
    }
  }
  ```
</Update>

<Update label="August 1, 2025">
  ## Private Beta Release MatterAI Axon private beta release
</Update>