MAI-Code-1.1-Flash
Microsoft's coding model supports agentic development in GitHub Copilot and downloadable local inference on high-memory hardware.
Mercury 3
www.inceptionlabs.ai
Inception's diffusion language model provides long-context text generation, adjustable reasoning, tool calling and structured outputs through an API.
Visit website
Pricing
Freemium
Available on
API
Category
Llm Models
What is Mercury 3?
Mercury 3 is Inception's flagship diffusion-based language model for text applications. It targets conversational products and agent workflows, including interactive education, search assistants and customer support.
Developers access Mercury through an OpenAI-compatible API. Its million-token chat context accommodates extensive input, while configurable reasoning lets applications choose how much deliberation happens before an answer.
Key features
Large context and output limits
Mercury 3 supports a one-million-token chat context and a separate maximum output allowance of 131,072 tokens.
Adjustable reasoning
Choose instant, low, medium or high reasoning effort. The default is low; higher settings allow deeper deliberation with greater latency.
Tool calling
The chat-completions endpoint supports tool calls for applications that connect model responses to external functions.
Structured outputs
Request structured responses for application workflows that need predictable output formats.
Compatible API integration
The OpenAI-compatible interface supports existing API integration patterns. The quickstart provides Python, TypeScript and curl examples.
Who is it for?
Education developers building tutors
Create conversational explanations and follow-up questions around supplied learning material.
Support teams building assistants
Combine support instructions with tool calls for account-specific workflows.
Developers processing extensive text
Evaluate long documents or project context within a conversational application.
How to use Mercury 3
- Create an Inception Platform account.
- Generate an API key and configure it for your development environment.
- Use the documented Python or TypeScript setup, or use curl.
- Send a chat-completions request using the mercury-3 model identifier.
- Evaluate the response, then adjust reasoning effort and add billing information when needed.
Mercury 3 pricing
Freemium
New accounts receive an introductory allocation of 100 million tokens without payment details. Continued API use is usage-based; enterprise arrangements are available through sales.
Check current pricing ↗What to know before you use it
- Mercury 3 is documented as a text model. Its supported-format listing does not include image or audio input.
- Temperature must be between 0.5 and 1.0, so integrations configured for temperature zero need adjustment.
- Higher reasoning effort increases latency; high is not the default setting.
- The free token allocation is introductory. Continued use beyond it requires billing information.
Frequently asked questions
Does the million-token context mean a million-token response?
No. The chat context is one million tokens, while the maximum output is 131,072 tokens.
Which reasoning setting should I start with?
Low is Mercury 3's default. Instant favors minimal latency, while medium and high provide progressively more deliberation.
Can I evaluate it before adding payment details?
Yes. The introductory token allocation is available without payment details after account and API-key setup.