Skip to content

feat(stepfun-ai): add step-5-preview - #7657

Merged
rekram1-node merged 1 commit into
anomalyco:devfrom
ArpitMishra17:feat/stepfun-ai-step-5-preview
Sep 21, 2026
Merged

rekram1-node merged 1 commit into
anomalyco:devfrom
ArpitMishra17:feat/stepfun-ai-step-5-preview

Conversation

@ArpitMishra17

Copy link
Copy Markdown
Contributor

Summary

Adds step-5-preview to the stepfun-ai (Global PAYG) provider, which serves it at api.stepfun.ai/v1 but has no catalog entry, so stepfun-ai/step-5-preview is unselectable in OpenCode. Reuses existing stepfun/step-5-preview lab metadata + base_model override pattern.

Data sources (all accessed 2026-09-21)

  • Model docs: https://platform.stepfun.ai/docs/en/guides/models/step-5-preview — 1M context/input, text/image/video in, tool calling, JSON Mode/Schema, reasoning effort low/medium/high (reasoning_effort on Chat, output_config.effort on Messages).
  • Pricing: https://platform.stepfun.ai/docs/en/guides/pricing/details — input $1.00, cache hit $0.05, output $2.70 per 1M.
  • Live API: GET https://api.stepfun.ai/v1/models lists step-5-preview with reasoning_effort_support_list: ["low","medium","high"], max_input_tokens: 1024000, supported_protocols: ["chat","messages","responses"]; POST /v1/chat/completions succeeds with a reasoning_content side channel.

Files

  • providers/stepfun-ai/models/step-5-preview.toml (new, override-only): effort low/medium/high, [interleaved] field = "reasoning_content", [cost] 1 / 2.7 / 0.05.
  • providers/stepfun-ai/provider.toml: effort comment extended to step-5-preview; Responses line updated (API reports responses support).

Judgment calls

  • Standalone file, not symlinked to providers/stepfun/: China (api.stepfun.com) returns 401 for a Global key, so parity is unverified — happy to convert to canonical+symlink if a maintainer confirms.
  • Not added to stepfun-ai-step-plan: Global Step Plan /models returns only 3.5 models for a working key.
  • Docs say max output 64k vs lab output = 1M: inherited lab value per feat(stepfun-step-plan): add step-5-preview #7544 precedent; flagging in case an override is preferred.
  • temperature/knowledge omitted (undocumented).

Validation

bun validate passes; resolved entry merges as expected.

StepFun's flagship base model, verified live on the Global PAYG gateway
(api.stepfun.ai/v1). Reuses existing lab metadata
(models/stepfun/step-5-preview.toml): 1M-token context/input, text/image/video
input, reasoning + tool calling + structured output, closed weights.
Provider override: reasoning_effort accepts low/medium/high; reasoning side
channel is reasoning_content; per-token cost input 1.00, output 2.70,
cache_read 0.05 USD/MTok.

Standalone file (not symlinked to providers/stepfun) because China
(api.stepfun.com) is unverified with a Global key (401). Global Step Plan
(step_plan/v1) verified absent for this model, so not added there.

Sources:
- https://platform.stepfun.ai/docs/en/guides/models/step-5-preview
- https://platform.stepfun.ai/docs/en/guides/pricing/details
- GET https://api.stepfun.ai/v1/models (accessed 2026-09-21)

bun validate passes.
@github-actions

Copy link
Copy Markdown
Contributor

No actionable findings.

@github-actions github-actions Bot added the reviewer: ready Automated review found no actionable items label Sep 21, 2026
@rekram1-node
rekram1-node merged commit a7332d5 into anomalyco:dev Sep 21, 2026
2 checks passed
@ArpitMishra17
ArpitMishra17 deleted the feat/stepfun-ai-step-5-preview branch September 23, 2026 00:32
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

reviewer: ready Automated review found no actionable items

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants