magma

Magma

magma

Magma 2.1

magma

Pyro

​
Magma

The most capable model layer with the largest context window and state-of-the-art video generation. Capabilities:
  • SOTA video generation
  • Function calling
  • Reasoning
  • Text-to-Speech (TTS)
  • Image-to-Image
  • Text/Image-to-Video
  • Video-to-Video
  • Text-to-Music / Music-to-Music
Best for: Long-context tasks, video generation, complex agentic workflows, and maximum quality output across all modalities.

​
Magma 2.1

Capabilities:
  • Strong agentic capabilities
  • Long-horizon task execution
  • Agentic coding
  • SOTA video generation
  • Function calling
  • Reasoning (always on; low / high / max effort)
  • Text-to-Speech (TTS)
  • Image-to-Image
  • Reference to Image
  • Text/Image-to-Video
  • Reference to Video
  • Video-to-Video
  • Text-to-Music / Music-to-Music
Highlights:
  • Token-efficient: stronger agentic coding results at every effort level while consuming fewer output tokens than the prior generation
  • Vulnerability discovery and security analysis
  • Full 1M context for whole-repository and long-document work
  • Max effort is recommended for coding tasks
Best for: Agentic workflows, long-horizon and multi-step coding tasks, reference-guided image and video generation, long-context tasks, and maximum quality output across all modalities.

​
Pyro

Capabilities:
  • SOTA agentic coding
  • Function calling
  • Structured output
  • Reasoning
Best for: Agentic coding, code generation and refactoring, tool-driven development workflows, and long-context codebase reasoning.

​
Comparison