Models

Autohand models

Fantail and Moa are Autohand's coding models. Weka is a typed decision model for release gates, operations, and agent workflows.

Choose a model

Model Model ID Use it for Context Max output Plans
Fantail fantail Completions, quick fixes, review comments, test scaffolds, and short agent loops. 256K input 16K Free and above
Moa moa Large refactors, architecture work, security reviews, and multi-file planning. 1M input 262K Pro and above
Weka weka Typed release, risk, triage, and workflow decisions with probabilities. 32K input Typed answer Pro and above

Fantail and Moa accept provider-qualified aliases as well as their bare identifiers. Weka uses the exact model ID weka at POST /v1/decisions. It is in Public Preview and available only through the Autohand API, Autohand SDK, and Console Playground. The CLI lists Weka for discovery but cannot run it.

Use a model

Pass the model identifier where you start the run. Use fantail for latency-sensitive work and moa when the task needs broad repository context or deliberate reasoning.

CLI

autohand --model fantail --prompt "Find the smallest safe fix for the failing test and apply it." --patch

# Choose Moa for repository-level planning
autohand --model moa --prompt "Map the authentication flow and propose the safest refactor plan." --patch

cURL

curl https://api.autohand.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AUTOHAND_API_KEY" \
  -d '{
    "model": "moa",
    "messages": [
      {"role": "user", "content": "Plan a safe migration for this service."}
    ]
  }'

Fantail

Fantail is the latency-first cloud model and the default selection for Autohand Code. Its 256K input context and tool calling support suit focused implementation loops.

  • Best for autocomplete, quick bug fixes, code review notes, documentation snippets, and small tests.
  • Works well when the relevant files are already in the prompt or easy for the agent to inspect.
  • Use Moa instead when the task depends on broad architecture or a large amount of repository context.

Read the Fantail model reference.

Moa

Moa is the reasoning-first cloud model for deeper software engineering work. Its 1M input context, tool calling support, and medium, high, and xhigh reasoning settings fit complex repository work.

  • Best for large refactors, migration plans, security audits, technical design, and repository analysis.
  • Use it when the model needs to compare patterns across a codebase before making changes.
  • Switch back to Fantail for final polish, small follow-up edits, and fast review cycles.

Read the Moa model reference.

Weka

Weka is trained on open source foundations, distilled from Fantail, and runs through Autohand inference as its first System One Model, trained with Reinforcement Learning for Calibrated Decisions (RLCD).

  • Best for release gates, incident triage, workflow routing, and agent checkpoints.
  • Use it when application code needs a validated value instead of generated prose.
  • Weka is in Public Preview and available only through the API, SDK, and Console Playground.
  • Weka returns typed noul, choice, and score answers with probabilities.
  • Use POST /v1/decisions with model ID weka; Weka does not use the chat-completions contract.

Read the Weka guide and tutorials.

Common patterns

  • Start fast. Use Fantail for the first pass when the task is scoped to a file, function, or failing test.
  • Escalate for context. Switch to Moa when the answer depends on architecture, cross-package contracts, or hidden coupling.
  • Make the branch explicit. Use Weka when the next step needs a typed choice, probability, or score.
  • Keep the model explicit. Include the model in scripts and SDK calls so runs are repeatable.