xMAD.ai

Models

The xMAD-2 family.

Four checkpoints, one training recipe. Released weights for the Mini and Base sizes; the Pro sizes are available to research partners under a capability agreement.

xMAD-2 MiniApache 2.0
Parameters
3B
Context
128k
AgentBench-2
48.1

Runs comfortably on a single 24 GB accelerator. Intended for evaluation work, on-device tool use, and as a distillation target.

xMAD-2 BaseApache 2.0
Parameters
9B
Context
256k
AgentBench-2
61.7

Our general-purpose open checkpoint. Strong tool-use calibration and the smallest model that reliably completes long multi-step tasks.

xMAD-2 ProxMAD Research
Parameters
52B
Context
512k
AgentBench-2
72.9

Frontier checkpoint, shared with research partners under a capability agreement. Requires a documented review of intended use.

xMAD-2 Pro LongxMAD Research
Parameters
52B
Context
2M
AgentBench-2
74.3

Trained for context rollover: state is summarised and re-expanded instead of truncated. Best results on tasks exceeding 500k tokens.

Model card — xMAD-2 Base

Every release ships with a card covering training data, evaluation, known limitations and the outcome of the pre-release review. The summary below is the short form; the full card is included with the weights.

Release date14 August 2026
Training dataPublic web text to March 2026, permissively licensed code, and 1.4M synthetic agent trajectories generated in simulation.
EvaluationAgentBench-2 61.7 · Terminal-2 54.9 · Tessera-information 0.62 · long-context retrieval 91.4% at 200k tokens.
Known limitationsDegrades on tasks requiring more than roughly 40 sequential tool calls. Over-uses the browser tool when a direct API call is available. Weak on non-Latin script tool outputs.
Pre-release reviewPassed. Two red-team findings were fixed before release; one — a prompt pattern that induced unsafe file deletion in sandboxed mode — is documented in section 6 of the full card.
LicenceApache 2.0 for weights and code. Use is subject to the Acceptable Use Policy.