Muse Spark 1.3 Free: free daily model for OpenCode

Introduction

Muse Spark 1.3 Free is the free contributor-tier variant of Muse Spark 1.3 from Meta, served through OpenCode Console. It is the recommended daily default: strong enough for explaining code, writing tests, small features, renames across files and docs, while costing nothing. The paid twin opencode/muse-spark-1.3 behaves the same when free quota runs out. Siblings in the trio: GPT 6 Luna for cheap paid work and Claude Opus 5.5 for the top final pass.

Pricing and ids

Endpoint family:

https://opencode.ai/inference/v1/responses

Live list with metadata:

https://opencode.ai/inference/v1/models

Prices move — verify in the Console models docs before budgeting. Free ids rotate; when one disappears pick another -free id from /models or fall back to local Ollama.

Connect in OpenCode

Connect Console once, then pin Spark Free as the project default:

opencode auth login opencode

opencode models

{ "$schema": "https://opencode.ai/config.json", "model": "opencode/muse-spark-1.3-contributor-free" }

Switch inside the TUI with /models or cycle recent models with F2. No billing details are needed for the free tier.

Skills and agents fit

Spark Free reads the same skill store as the other models but follows short imperative skills best: one screen, numbered steps, exact commands, one example. Best matches from Five complete examples: commit-style and test-writer. In the aredel repo it owns aredel-map only. Agent side, it runs triage: read-only, capped steps, no edits, routes work to builder or plan. Keep long block rules in references, not in the hot skill body.

Example workflow

Morning triage on Spark Free, escalate only what needs tools:

With triage, explain src/auth.ts login flow and route the refresh bug as small fix or feature

@test-writer Cover src/auth.ts with three tests

If the answer is weak, stay in the session and hand the exact context to GPT 6 Luna for the build step and to Claude Opus 5.5 for the verdict. This split keeps roughly 80 percent of steps free.

Privacy and limits

Contributor-tier privacy: the free variant uses heavily discounted pricing in exchange for permission to use prompts and completions to train future models. Never send secrets, customer data, or private code to it. Most other free models claim zero-retention during the free period, but some may still use data for improvement — check the Console privacy notes first, and switch to a paid model or a local model for sensitive work.

Free tier is rate-limited and time-limited for feedback collection. For overnight runs and CI use GPT 6 Luna instead.

Troubleshooting

• Free model gone: free ids rotate — run /models, pick another -free id or the paid twin, or fall back to Ollama.
• Weak answers: add file context with @, shorten the skill to one screen, move policy prose to references.
• Rate-limited mid-session: keep context, switch model with F2 to GPT 6 Luna, continue without re-reading files.

Official sources:

• opencode.ai/v2/docs
• opencode.ai/docs/console/models

Article author: Arthur Isaev