KPI Snapshot
2
Agent Models
Hunter Alpha · Healer Alpha
1T
Hunter Parameters
Largest dense model on OpenRouter
4
Healer Modalities
Text · Image · Audio · Video
$0
Pricing
All token types — alpha access
Spec Comparison
Core Specifications
Hunter Alpha
Healer Alpha
Parameters
1 Trillion
Undisclosed
Context window
1,048,576
262,144
Max completion
32,000
32,000
Modalities
Text → Text
Text+Img+Audio+Video → Text
Tool calling
✓
✓
Reasoning tokens
✓
✓
Structured output
✓
✓
Default temp
1.0
1.0
Pricing
Free
Free
Data logging
All I/O logged
All I/O logged
Capability Speed Bars
Context Window
Input Modalities
Parameter Scale (where disclosed)
Agent Parameter Support
API Parameters (both models identical)
tools
✓ Supported
✓ Supported
tool_choice
✓ Supported
✓ Supported
include_reasoning
✓ Supported
✓ Supported
response_format
✓ Supported
✓ Supported
frequency_penalty
✓ Supported
✓ Supported
presence_penalty
✓ Supported
✓ Supported
stop sequences
✓ Supported
✓ Supported
logprobs
✗ Not supported
✗ Not supported
seed
✗ Not supported
✗ Not supported
Design Philosophy
Hunter
Role
Deep reasoning, long-horizon planning, multi-step execution chains
Healer
Role
Multi-modal perception, cross-modal reasoning, real-world sensing
Open Questions
Benchmarks
Status
No published scores on SWE-Bench, AgentBench, or standard evals
Provider
Status
OpenRouter namespace, but actual trainer undisclosed
Latency
Status
1T params = expensive serving; real-world latency unreported
Timeline
Status
Alpha → production path unknown; pricing TBD
Data: OpenRouter API (/api/v1/models), queried March 14, 2026. All specs verified against live API response. Both models created ~March 9, 2026.