Quantel AI Research · Multi-Core Language Models
Read the paper →
Quantum model catalog

Models built for different engineering loops.

Choose the model profile that fits your workload. Both available Quantum models support a 1M-token context window, 128K-token output and vision input.

Available now1M context128K outputVision input
Three native interfaces

Use OpenAI Responses, OpenAI Chat Completions or Anthropic Messages with the model profile that fits your application.

Available models

Pick the right balance of speed and depth.

Quantum Lite 1 is designed for quick interaction. Quantum Ultra 1 is the production flagship for reasoning-intensive software engineering.

Fast feedback loops

Quantum Lite 1

model: quantum-lite

Quantum series · Available now

A direct, low-latency model profile for short coding loops, CLI assistants and high-volume interactive requests.

1,000,000 tokens
128,000 tokens
Supported
25 August 2026
January 2026
Single-call profile
SSE-compatible streaming
Chat Completions · Responses · Anthropic Messages
Quantum Lite architecture
Quantel Quantum serving stack

Best for

  • CLI and terminal assistants
  • Rapid code edits
  • Interactive product features

Product notes

  • Model behavior varies with request complexity and serving conditions.
  • Image inputs are supported for visual coding and analysis workflows.
Join for quantum-liteView interfaces
Production flagship

Quantum Ultra 1

model: quantum-ultra

Quantum series · Available now

A reasoning-oriented model for agentic software engineering, visual analysis and native streaming workflows.

1,000,000 tokens
128,000 tokens
Supported
25 August 2026
January 2026
Single-call native-SSE profile
Native SSE streaming
Chat Completions · Responses · Anthropic Messages
MCLM / QSA architecture
Quantel Quantum serving stack

Best for

  • Agentic software engineering
  • Repository analysis
  • Tool-using coding workflows

Product notes

  • Reasoning effort defaults to max in the gateway contract.
  • Throughput varies with workload, context, provider and serving conditions.
  • Image inputs are supported for visual coding and analysis workflows.
  • Published benchmark figures are Quantel internal evaluations, not official third-party leaderboard results.
Join for quantum-ultraView interfaces
At a glance

Compare the available Quantum models.

Both models share the same core public specifications while targeting different engineering workloads.

SpecificationQuantum Lite 1Quantum Ultra 1
Model IDquantum-litequantum-ultra
Context window1,000,000 tokens1,000,000 tokens
Maximum output128,000 tokens128,000 tokens
Vision inputSupportedSupported
Release date25 August 202625 August 2026
Knowledge cutoffJanuary 2026January 2026
Best forFast feedback loopsAgentic engineering
ArchitectureQuantum Lite architectureMCLM / QSA architecture
Supported APIsResponses · Chat · AnthropicResponses · Chat · Anthropic
Public benchmarkNot publishedInternal evaluation only
Model ID
quantum-litequantum-ultra
Context window
1,000,000 tokens1,000,000 tokens
Maximum output
128,000 tokens128,000 tokens
Vision input
SupportedSupported
Release date
25 August 202625 August 2026
Knowledge cutoff
January 2026January 2026
Best for
Fast feedback loopsAgentic engineering
Architecture
Quantum Lite architectureMCLM / QSA architecture
Supported APIs
Responses · Chat · AnthropicResponses · Chat · Anthropic
Public benchmark
Not publishedInternal evaluation only
Three native interfaces

Use the client your stack already understands.

Select a Quantum model through the supported protocol interface that fits your tooling. Join the waitlist for API access and launch updates.

  • OpenAI Responses for Codex and agents
  • OpenAI Chat Completions for existing SDKs
  • Anthropic Messages for Claude-compatible clients
POST /v1/responses{"model": "quantum-ultra","input": "Review this repository","stream": true}
What comes next

Aura and Avatar are future families, not current API models.

The roadmap is intentionally separate from the available catalog. Future names do not imply a launch date, public model ID or production availability.

Future · not available

Aura series

A future family of highly specialized MCLMs intended for demanding domains such as cybersecurity, biology and defense-related research.

Future · not available

Avatar series

A future family for image and video generation. No public model IDs, launch dates, pricing or production specifications have been published.

Build with Quantum

Choose the model profile for your next engineering loop.

Join the waitlist for early access to the Quntel model series and developer platform.

Join the waitlistRead the research