Agents-A1 Models by Intern Science

Intern ScienceApache 2.0Open sourceOpen SourceAgentsVision
Intern Science releases · 2 in the last 12 monthsChangelog →
2 models2026Up to 256K ctx

Last refreshed 2026-10-22. Next refresh: weekly.

Details

ResearcherIntern Science
LicenseApache 2.0OSI-approved
Commercial usePermitted
Models2
Released2026
Max context256K

Capabilities

VisionAll models
MultimodalAll models
ReasoningAll models
JSON / Tool useAll models

About

Agents-A1 is Intern Science's open-source 35B-A3B Mixture-of-Experts family for long-horizon agentic work across search, engineering, scientific research, instruction following, and tool calling.

Decision facts

Best fit
agentsagenticvision and multimodal work
Capability starting point
Sage-A1.5 with 256K context and reasoning, JSON / Tool use, and multimodal inputs
Lowest tracked input
Not tracked
Closest related family
Bonsai

Current Variants

Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.

2 in view
Sage-A1.5Current

Use when the workload needs agentic, 256K context, and reasoning.

2026-10agentic256K contextreasoning
Agents-A1Current

Use when the workload needs 256K context, reasoning, and JSON / Tool use.

2026-06256K contextreasoningJSON / Tool use

Release Timeline

2 release groups
2026-10
1 current
Sage-A1.5
agentic256K contextreasoning
Current
2026-06
1 current
Agents-A1
256K contextreasoningJSON / Tool use
Current

Specifications(2 models)

Agents-A1 model specifications comparison
ModelReleasedContextParametersVisionMultimodalReasoningJSON / Tool use
Sage-A1.52026-10256K35B total, 3B activeYesYesYesYes
Agents-A12026-06256K35B total, 3B activeYesYesYesYes