Muse Spark 1.3 in OpenCode: id, pricing and the retired free tier

Introduction

Muse Spark 1.3 from Meta is served through OpenCode Console as a paid model, opencode/muse-spark-1.3. Until October 2026 it also had a free contributor-tier variant, Muse Spark 1.3 Free, which these guides recommended as the daily default. That variant is no longer in the catalog; its story is under History. For a free model today use the free OpenCode model, for cheap paid work GPT 6 Luna, and for the top final pass Claude Opus 5.5.

Pricing and ids

Endpoint family:

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

Live list of ids:

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

Prices move — verify in the Console models docs before budgeting.

Connect in OpenCode

Connect Console once and add credits:

opencode auth login opencode

opencode models

Then pin the paid id in opencode.jsonc at the project root:

vi opencode.jsonc

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

Switch inside the TUI with /models or cycle recent models with F2.

Skills and agents fit

Muse Spark 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. Keep long block rules in references, not in the hot skill body. The read-only triage role now runs on the free OpenCode model.

When it is worth it

At $1.25 in and $4.25 out per 1M tokens, Muse Spark 1.3 costs about twelve times as much as GPT 6 Luna on input. Pick it when a skill or prompt was tuned on Muse Spark 1.3 Free and must keep the same behavior on a stable paid quota. For everything else Luna is the cheaper paid step and the free OpenCode model the free one.

History: Muse Spark 1.3 Free

Muse Spark 1.3 Free, id opencode/muse-spark-1.3-contributor-free, was the free contributor-tier variant of Muse Spark 1.3, served through OpenCode Console. It was 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 behaved the same when free quota ran out.

Contributor-tier privacy: the free variant used heavily discounted pricing in exchange for permission to use prompts and completions to train future models. Secrets, customer data and private code were never to be sent to it. The same rule holds for any free model today: check the Console privacy notes for the exact id first.

The free tier was rate-limited and time-limited for feedback collection. The workflow it supported was a free morning triage with escalation:

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

Weak answers went to GPT 6 Luna for the build step and to Claude Opus 5.5 for the verdict, which kept roughly 80 percent of steps free. On 2026-10-22 the models endpoint no longer listed the free id. The same workflow now starts on the free OpenCode model; that page also explains why free ids should be checked before they are pinned.

Troubleshooting

• Model error on start with the free id: opencode/muse-spark-1.3-contributor-free is retired; pin the free OpenCode model or the paid opencode/muse-spark-1.3.
• Weak answers: add file context with @, shorten the skill to one screen, move policy prose to references.
• Bill higher than expected: move routine steps to GPT 6 Luna, check opencode stats --cost --models.

Official sources:

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

Article author: Arthur Isaev