Name of model:
- LongCat 2.5 Preview Free — longcat-2.5-preview-free.
- Step 5 Preview Free — step-5-preview-free.
- Space Bunny — space-bunny.
Model creator: Not established by the OpenCode Go sources consulted for these gateway entries. Creator attribution must be confirmed before adding descriptions; in particular, this request does not infer an underlying vendor for Space Bunny.
Link to model: OpenCode Go lineup, official Go model/endpoint table, Go pricing, and live Go model catalog.
Problem and user benefit
All three IDs are listed by OpenCode Go and discovered through its catalog, but none has a native Zoo Code Go configuration. Users therefore get generic context/output limits and capability flags, with unknown pricing even for models advertised as free.
Add confirmed configurations for this small batch of newly listed Go models, so their settings and displayed costs match the gateway's actual offerings. This is a single catalog-completion task; it does not assert that their existing Chat Completions route fails.
Concrete evidence
Audited upstream main at ecab5198, October 10, 2026. Relevant sources and tests were verified byte-identical to the locally executed checkout at 50e7aea3.
- Offline execution of the native Go metadata lookup returned no entry for each of the three IDs.
- The live catalog observed during this audit returned IDs and administrative fields only. The fetcher's unknown-model branch consequently supplies generic 200,000-token context, 32,768-token output, disabled image/prompt-cache capability flags, and no curated pricing/reasoning controls.
- The catalog-completeness test excludes all three IDs.
- The official Go endpoint table assigns all three to Chat Completions, which matches current routing. The missing work is model configuration, not protocol migration.
Desired behavior and acceptance criteria
Scope and related work
Keep this task limited to the three missing Go model configurations. It does not introduce a new provider, redesign discovery, or automatically guess metadata for arbitrary future IDs.
Searched open and closed issues/PRs by the model IDs/display names and OpenCode Go; no direct duplicate was found. The separate MiMo V2.6 request concerns vendor-specific reasoning continuity and is intentionally outside this small generic catalog batch. PR #1988 concerns GPT Responses routing and does not configure these IDs.
Verification performed
Direct offline lookup confirmed all three models are uncurated. Existing focused registry/provider/fetcher suites passed: 41 + 93 tests. This is baseline verification, not regression coverage for the missing models. No authenticated inference requests or measurements of model limits/capabilities were performed.
Name of model:
Model creator: Not established by the OpenCode Go sources consulted for these gateway entries. Creator attribution must be confirmed before adding descriptions; in particular, this request does not infer an underlying vendor for Space Bunny.
Link to model: OpenCode Go lineup, official Go model/endpoint table, Go pricing, and live Go model catalog.
Problem and user benefit
All three IDs are listed by OpenCode Go and discovered through its catalog, but none has a native Zoo Code Go configuration. Users therefore get generic context/output limits and capability flags, with unknown pricing even for models advertised as free.
Add confirmed configurations for this small batch of newly listed Go models, so their settings and displayed costs match the gateway's actual offerings. This is a single catalog-completion task; it does not assert that their existing Chat Completions route fails.
Concrete evidence
Audited upstream main at ecab5198, October 10, 2026. Relevant sources and tests were verified byte-identical to the locally executed checkout at 50e7aea3.
Desired behavior and acceptance criteria
Scope and related work
Keep this task limited to the three missing Go model configurations. It does not introduce a new provider, redesign discovery, or automatically guess metadata for arbitrary future IDs.
Searched open and closed issues/PRs by the model IDs/display names and OpenCode Go; no direct duplicate was found. The separate MiMo V2.6 request concerns vendor-specific reasoning continuity and is intentionally outside this small generic catalog batch. PR #1988 concerns GPT Responses routing and does not configure these IDs.
Verification performed
Direct offline lookup confirmed all three models are uncurated. Existing focused registry/provider/fetcher suites passed: 41 + 93 tests. This is baseline verification, not regression coverage for the missing models. No authenticated inference requests or measurements of model limits/capabilities were performed.