Add Step 3.7 Flash and StepFun Step Plan providers by lit26 · Pull Request #1903 · anomalyco/models.dev

lit26 · 2026-05-29T17:37:37Z

Summary

Add step-3.7-flash.toml + logo.svg for stepfun and stepfun-ai providers
Add stepfun-step-plan provider (api.stepfun.com/step_plan/v1) — step-plan reasoning API (zh)
Add stepfun-ai-step-plan provider (api.stepfun.ai/step_plan/v1) — step-plan reasoning API (en)
Fix stepfun-ai to use standard api.stepfun.ai/v1 endpoint + overview doc (was incorrectly on step_plan/v1)
Set reasoning = false on step-1-32k / step-2-16k (legacy chat models, not reasoning models)

Step Plan providers

Step-plan reasoning API (refs: zh, en). Supports reasoning_effort low/medium/high.

Models:

step-3.7-flash, step-3.5-flash, step-3.5-flash-2603 — in both providers
step-router-v1 — stepfun-step-plan (.com) only

All step-plan models omit [cost]: billing is plan-based, not per-token.

Step 3.7 Flash details

Released 2026-05-29, open weights
Multimodal input: text / image / video; text output (video declared on the base model in models/stepfun/step-3.7-flash.toml)
256K context / input / output
Reasoning, tool calling, temperature, attachments

Sources / citations

Multimodal (image + video) input — StepFun reasoning-models guide: "新增原生多模态输入能力，可直接理解图片和视频内容" (zh, en).
Reasoning models list — the same guide lists only Step 3.7 Flash / Step 3.5 Flash (2603) as reasoning models. step-1-32k and step-2-16k are legacy non-reasoning chat models and are absent from the reasoning list, hence reasoning = false (they were incorrectly reasoning = true on dev).
China per-token cost — StepFun pricing details (zh): step-3.7-flash input (cache miss) ¥1.35/1M, output ¥8.1/1M, input (cache hit) ¥0.27/1M. TOML values 0.185 / 1.11 / 0.037 are the USD conversions.
Global per-token cost — $0.20 in / $1.15 out / $0.04 cache read per 1M (standard api.stepfun.ai/v1).
Knowledge cutoff 2026-03-01 — StepFun publishes no official cutoff; this remains an estimate from the model's release/training timeline. Will revise if an authoritative source surfaces.

Test plan

bun validate passes (EXIT 0); all providers load
New providers render with logo and correct endpoints

Add step-3.7-flash.toml for both providers (released 2026-05-29, multimodal text/image/video input, 256K context, open weights) plus provider logos.

Add stepfun-step-plan (api.stepfun.com/step_plan/v1) and stepfun-ai-step-plan (api.stepfun.ai/step_plan/v1) providers for the StepFun step-plan reasoning API. Models: step-3.7-flash, step-3.5-flash, step-3.5-flash-2603 in both; step-router-v1 in the .com provider only. All step-plan models omit cost (plan-based billing). Fix stepfun-ai provider to point at the standard api.stepfun.ai/v1 endpoint and overview doc, separating it cleanly from the step-plan API.

hengm3467 · 2026-05-30T02:48:52Z

@lit26
stepfun and stepfun-ai might not be the best provider names (might confuse people) - I suggest "StepFun (China)" "StepFun (Global)“

StepRouter might be mature enough to be added to the catalog. Need confirmation with PM (Bingzhe).
providers/stepfun/models/step-3.7-flash.toml (line 12): the China provider copies the global USD pricing 0.20 / 1.15 / 0.04, but StepFun China docs list ¥1.35 / ¥8.1 / ¥0.27. This repo stores costs in USD, and existing step-3.5-flash converts the China RMB price rather than copying the global .ai price, so these should likely be around 0.185 / 1.11 / 0.037.

- Convert step-3.7-flash China cost from copied global USD to RMB-converted USD (0.185/1.11/0.037 from ¥1.35/8.1/0.27), matching step-3.5-flash convention - Rename providers to disambiguate China vs Global variants

lit26 · 2026-05-30T05:11:54Z

Thanks @hengm3467! Addressed in 467363a:

Pricing: fixed providers/stepfun/models/step-3.7-flash.toml — the China provider was copying the global USD price. Converted from China docs ¥1.35 / 8.1 / 0.27 → 0.185 / 1.11 / 0.037 USD, matching the step-3.5-flash RMB-conversion convention. Global (stepfun-ai) stays 0.20 / 1.15 / 0.04.
Naming: renamed for China/Global symmetry — StepFun (China) / StepFun Step Plan (China) alongside the existing (Global) variants.

rekram1-node · 2026-06-13T19:22:03Z

AUTOMATED REVIEW: This PR is not yet compliant with the model/provider contribution requirements.

Please use the available StepFun canonical metadata via base_model for the added step-3.5-flash-2603, step-3.5-flash, and step-3.7-flash files across providers/stepfun-ai-step-plan/, providers/stepfun-ai/, providers/stepfun-step-plan/, and providers/stepfun/.

All eight of those reasoning model files must explicitly specify reasoning_options; providers/stepfun-step-plan/models/step-router-v1.toml must do so as well.

The logos for the two new providers, providers/stepfun-ai-step-plan/logo.svg and providers/stepfun-step-plan/logo.svg, have fixed width="28" and height="28". Per the logo guidelines, new provider SVGs must have no fixed size or colors and must use currentColor for fills/strokes.

- Collapse stepfun/stepfun-ai/*-step-plan model files to base_model refs - Add reasoning_options (effort levels) per model - Add benchmarks and update knowledge date on base step-3.7-flash - Normalize logo.svg viewBox

lit26 · 2026-06-14T18:01:02Z

Pushed an update (f55836a):

Refactored the StepFun provider model files (stepfun, stepfun-ai, stepfun-step-plan, stepfun-ai-step-plan) to inherit from the base stepfun/* models via base_model, removing duplicated name/limit/modalities/flags.
Added reasoning_options (effort levels) per model.
Added benchmarks (Artificial Analysis Coding Index, SciCode, Terminal-Bench Hard) and updated the knowledge cutoff to 2026-03-01 on the base step-3.7-flash.
Normalized logo.svg to use viewBox only.

javlund · 2026-06-27T11:40:44Z

Hoping that this will be merged soon - it's annoying that 3.7 isn't available without config changes in Opencode.

…d-step-3.7-flash # Conflicts: # providers/stepfun-ai/provider.toml # providers/stepfun/models/step-3.7-flash.toml

Merge resolution dropped the reasoning-format documentation comments from stepfun-ai/provider.toml. Restore them (mirroring the China stepfun/provider.toml block, adapted to the global .ai /v1 endpoints) per reviewer request on PR anomalyco#1903.

… comments

Output was artificially capped at 65536; StepFun docs set max_tokens default to INF, bounded only by the 256K context window. Raise step-3.7-flash maxTokens to 262144 on both stepfun and stepfun-plan, matching contextWindow. Per models.dev anomalyco/models.dev#1903. step-3.5-flash entries unchanged.

Step Plan endpoint bills per subscription plan, not per token, matching the sibling stepfun-plan/step-3.5-flash and step-3.5-flash-2603 entries which already carry zero cost. Standard stepfun/step-3.7-flash keeps its per-token rate. Per models.dev anomalyco/models.dev#1903, whose step-plan providers omit cost entirely.

# Conflicts: # providers/stepfun/models/step-3.7-flash.toml

# Conflicts: # providers/stepfun-ai/models/step-3.7-flash.toml # providers/stepfun-ai/models/step-3.7-flash.toml~HEAD

rekram1-node · 2026-07-07T23:18:26Z

+# POST /v1/chat/completions uses top-level `reasoning_effort`; POST /v1/messages
+# uses `output_config.effort`; POST /v1/responses uses `reasoning.effort`.
+# Values are low/medium/high for step-3.7-flash; step-3.5-flash-2603 accepts
+# low/high. Responses supports only step-3.7-flash. Chat returns reasoning at
+# `choices[].message.reasoning` or streamed `choices[].delta.reasoning`;
+# `reasoning_format` is general (default), the latter using `reasoning_content`.


the comments document effort as being supported but none of the models specify reasoning options correctly if that is the case?

The docs match now: effort reasoning_options are declared for the actual reasoning models across all four providers — step-3.7-flash (low/medium/high), step-3.5-flash and step-3.5-flash-2603 (low/high). step-1-32k/step-2-16k are legacy Step 1/Step 2 chat models and step-router-v1 is a dispatcher — none are reasoning models per StepFun's docs, so they're now reasoning = false rather than reasoning = true with empty options

* feat(stepfun): add step-3.7-flash model and make it default * fix(stepfun): set step-3.7-flash cacheRead cost to 0.04 * fix(stepfun): raise step-3.7-flash maxTokens to context limit Output was artificially capped at 65536; StepFun docs set max_tokens default to INF, bounded only by the 256K context window. Raise step-3.7-flash maxTokens to 262144 on both stepfun and stepfun-plan, matching contextWindow. Per models.dev anomalyco/models.dev#1903. step-3.5-flash entries unchanged. * fix(stepfun): zero step-plan step-3.7-flash cost for plan billing Step Plan endpoint bills per subscription plan, not per token, matching the sibling stepfun-plan/step-3.5-flash and step-3.5-flash-2603 entries which already carry zero cost. Standard stepfun/step-3.7-flash keeps its per-token rate. Per models.dev anomalyco/models.dev#1903, whose step-plan providers omit cost entirely. * docs(changelog): note StepFun 3.7 support * style(stepfun): format provider table * fix(stepfun): forward reasoning effort * fix(stepfun): scope reasoning compat to 3.7 * test(stepfun): keep 3.5 reasoning compat unchanged * fix(stepfun): carry off fallback through agent turns * fix(stepfun): use documented token limit field * fix(stepfun): match chat completions contract * chore(stepfun): keep changelog release-owned --------- Co-authored-by: Tianning Li <[email protected]> Co-authored-by: Peter Steinberger <[email protected]>

Step 1/Step 2 are legacy chat models, not reasoning models, so drop the placeholder reasoning=true/reasoning_options=[]. Give step-3.5-flash the effort low/high options its 2603 snapshot already declares, matching the provider reasoning-HTTP-format comment. Symlinked into stepfun-ai.

…easoning step-3.5-flash on both step-plan providers now declares effort low/high, matching the stepfun/stepfun-ai variants and its own 2603 snapshot. Step Router v1 is a dispatcher, not a reasoning model (per StepFun docs), so reasoning=false.

github-actions · 2026-07-09T09:28:41Z

Here is my review:

Review: PR #1903 — Add Step 3.7 Flash and StepFun Step Plan providers

Hard blockers: None ✅

All hard blockers from the contribution checklist are satisfied:

Logos shipped for all providers — stepfun-step-plan, stepfun-ai-step-plan (new) and stepfun, stepfun-ai (existing, previously missing) all get logo.svg with currentColor path fills and square viewBox="0 0 28 28".
base_model used when models/ metadata exists — step-3.7-flash, step-3.5-flash, and step-3.5-flash-2603 in both step-plan providers correctly reference base_model = "stepfun/..." instead of duplicating provider-agnostic facts. step-router-v1 uses an inline definition (no matching models/stepfun/step-router-v1.toml), which is correct.
reasoning_options declared for all reasoning models — step-3.7-flash (effort low/medium/high), step-3.5-flash and step-3.5-flash-2603 (effort low/high) all have options matching the documented API surface. Patches 8–9 correctly reconcile inconsistent options across variants.
Non-reasoning models cleaned up — step-1-32k, step-2-16k, and step-router-v1 correctly flip to reasoning = false with reasoning_options removed, satisfying the schema's refineModel constraints.

Non-blocking recommendations

Video modality should be in base model metadata, not provider overrides. The PR body describes video as a model capability ("Multimodal input: text / image / video"), but models/stepfun/step-3.7-flash.toml still has input = ["text", "image"]. Every provider TOML individually overrides to ["text", "image", "video"]. Adding "video" to the base model's [modalities].input would eliminate the repeated overrides and keep the model-level facts in models/.
Missing [interleaved] on new provider models. The current dev stepfun provider declares [interleaved] field = "reasoning_content" on step-3.5-flash, step-3.5-flash-2603, and step-3.7-flash. Since interleaved is provider-specific (not inherited via base_model), the new stepfun-ai and step-plan provider TOMLs should declare it too if the API returns interleaved reasoning. This is likely a rebase artifact — the PR's base predates these additions on dev.
Step-plan provider.toml files lack reasoning HTTP format comments. 54+ providers in the repo include # Reasoning HTTP format comment blocks. The PR adds them to stepfun-ai/provider.toml (patches 6–7) but not to the new stepfun-step-plan/provider.toml or stepfun-ai-step-plan/provider.toml. The step-plan endpoint's reasoning format should be documented there for consistency.
Comments in model TOMLs are between sections, not at the top. Reasoning-effort comments in providers/stepfun/models/step-3.5-flash.toml and step-3.7-flash.toml sit above [[reasoning_options]] rather than at the file top. Per AGENTS.md, comments should live at the top to survive sync re-serialization. Stepfun isn't currently synced, so this is a future risk.
Logo minor cleanup. The SVG root has fill="none" (harmless — the path overrides with fill="currentColor" — but unnecessary), the <clipPath> rect has an irrelevant fill="currentColor" (clip paths use geometry, not fill), and files lack a trailing newline.
PR body mislabels cost units. The body says "Cost per 1K tokens" but the schema and all other providers use per-million-token pricing. The TOML values (0.20/1.15/0.04 global, 0.185/1.11/0.037 China) appear correct as per-million prices; the description is just misleading.
Naming convention is inverted from repo standard. The repo convention is: base ID = global (alibaba → "Alibaba"), -cn ID = China (alibaba-cn → "Alibaba (China)"). This PR has stepfun (base ID) = China → "StepFun (China)" and stepfun-ai (suffixed ID) = Global → "StepFun (Global)". This is a pre-existing issue from original provider creation; the "(China)"/"(Global)" suffixes are reasonable given the constraint.
Merge conflict risk — rebase needed. dev has diverged significantly from the PR's base. stepfun-ai/provider.toml on dev now has name = "StepFun AI" with step-plan reasoning comments and [interleaved] declarations exist on stepfun model TOMLs. A rebase will be required; care should be taken to preserve dev improvements (especially [interleaved]) that the PR's diff doesn't account for.
Knowledge date change lacks citation. models/stepfun/step-3.7-flash.toml knowledge changes from "2026-01-01" to "2026-03-01" without a cited source in the PR body.
stepfun-ai only gets step-3.7-flash. Other StepFun models available on the global API (step-3.5-flash, step-3.5-flash-2603) are not added. Not a blocker for this PR, but worth noting for completeness.

Summary

Well-structured PR. The base_model refactoring, reasoning_options reconciliation (patches 8–9), and the clean separation of standard vs step-plan APIs are all done correctly. No hard blockers — the recommendations above are non-blocking improvements, with the rebase (item 8) being the most important practical concern before merge.

- Move video modality to base model step-3.7-flash; drop redundant per-provider [modalities] overrides (all 4 providers inherit video) - Add # Reasoning HTTP format comment to both step-plan provider.toml - Move reasoning-effort comments to top of model TOMLs for sync safety - Logo cleanup: drop root fill=none, clipPath rect fill; add trailing newline

github-actions · 2026-07-09T10:00:08Z

Reviewed the PR metadata, diff, AGENTS.md, README, schema, sync.md, and the reasoning-options audit guide, then reconstructed the final state of each changed file against the base checkout.

No blockers found: the four stepfun providers all ship compliant logos (square viewBox, currentColor, no fixed size), every new model with a matching models/stepfun/* entry uses base_model, and every reasoning = true model declares reasoning_options (with reasoning = false models correctly omitting it). The new stepfun-step-plan / stepfun-ai-step-plan providers and the stepfun-ai endpoint fix are structurally sound.

[low] providers/stepfun/models/step-3.7-flash.toml:861 (and providers/stepfun-ai/models/step-3.7-flash.toml:64) - The step-3.7-flash per-token costs (Global 0.20/1.15/0.04 and the China RMB-converted 0.185/1.11/0.037) are asserted in the PR body without a direct pricing-page citation; the PR links the Step Plan reasoning API but not a pricing source for these values. Add a first-party pricing link stating what each number supports.
[low] providers/stepfun/models/step-1-32k.toml:963 and providers/stepfun/models/step-2-16k.toml:977 - reasoning is flipped from true to false on existing models with no supporting citation in the PR body. This is a material capability change; add a source (model docs or API reference) confirming Step 1/Step 2 are non-reasoning chat models.
[low] providers/stepfun-ai/models/step-3.7-flash.toml:683 - This Global model omits [interleaved], while the China counterpart declares interleaved.field = "reasoning_content" and the provider's own reasoning-format comment documents a reasoning_content (deepseek-style) output. Confirm whether the Global /v1 endpoint intentionally lacks interleaved reasoning output; if not, add the field for parity with the China variant.
[low] providers/stepfun-ai/provider.toml:935 - The reasoning-format comment was left self-contradictory after the edit: "reasoning_format is general (default), the latter using reasoning_content" — "the latter" no longer has an antecedent (the "or deepseek-style" clause was removed), so the line now incorrectly implies the general/default format uses reasoning_content. Restore the antecedent or rewrite the sentence.

# Conflicts: # models/stepfun/step-3.7-flash.toml

Add [interleaved] reasoning_content field to all step-plan model files and normalize to multi-line TOML form matching the direct-API providers. Step-plan comment omits the Responses endpoint clause (no plan Responses endpoint per provider docs).

into add-step-3.7-flash

github-actions · 2026-07-11T05:20:47Z

Action items

[medium] [possible mistake] providers/stepfun-ai/models/step-3.7-flash.toml - Check: Provider-specific interleaved field for a reasoning model. Why: interleaved is not inherited via base_model (it is not part of model metadata), so it must be provider-authored. The sibling stepfun (China), stepfun-step-plan, and stepfun-ai-step-plan step-3.7-flash entries all declare [interleaved] field = "reasoning_content", and the stepfun-ai provider.toml reasoning comment references reasoning_content. This file omits it, so the generated Global model lacks the interleaved field while the China/step-plan variants expose it. Action: Add [interleaved] with field = "reasoning_content".
[medium] [possible mistake] models/stepfun/step-3.7-flash.toml - Check: Benchmark additions against the current base. Why: The PR appends three OpenRouter-sourced benchmarks (diff patch 4) assuming the file ends after the [[weights]] block, but the current dev base already has eight static.stepfun.com benchmark entries after that block. The diff context no longer matches the base, so a merge conflicts at that insertion point and risks dropping dev's existing eight benchmarks if resolved incorrectly. Action: Rebase onto current dev and reconcile so the three new benchmarks are added alongside, not replacing, the existing eight.
[low] [violation] providers/stepfun-ai/models/step-3.7-flash.toml - Check: Redundant provider-agnostic modality override. Why: After this PR moves video onto the base model (models/stepfun/step-3.7-flash.toml modalities.input), this file's [modalities] input = ["text", "image", "video"] is identical to the inherited value. The PR's own commit 10 states it drops redundant per-provider [modalities] overrides for "all 4 providers" but this file was missed. Action: Remove the [modalities] block so the value is inherited.
[low] [possible mistake] providers/stepfun/models/step-1-32k.toml, providers/stepfun/models/step-2-16k.toml - Check: Reasoning capability change. Why: reasoning is changed from true to false (and reasoning_options removed) with no cited source, contradicting the existing reasoning = true / reasoning_options = [] data. This is a material capability downgrade that cannot be verified from the PR. Action: Cite Step 1 / Step 2 model documentation proving they are non-reasoning, or revert.
[low] [possible mistake] providers/stepfun/models/step-3.7-flash.toml - Check: China per-token pricing citation. Why: The China cost is changed to 0.185 / 1.11 / 0.037 (described as an RMB conversion of ¥1.35 / 8.1 / 0.27) but no direct link to the StepFun China pricing page is provided, so the RMB rates cannot be verified. Action: Cite the China pricing page that states the per-token RMB rates.
[low] [possible mistake] models/stepfun/step-3.7-flash.toml - Check: Citations for the video modality and knowledge cutoff. Why: modalities.input gains video and knowledge changes to 2026-03-01 (acknowledged in the PR body as an estimate). Neither change cites a model card or API reference. Action: Cite the model card / API reference supporting video input and the knowledge cutoff, or revert to the prior values.

Global step-plan model docs now point to the .ai/en guides instead of the China .com/zh pages. Links verified live.

lit26 · 2026-07-11T14:05:33Z

Thanks for the review — status on each action item:

Resolved in the current branch

[medium] stepfun-ai step-3.7-flash missing [interleaved] — providers/stepfun-ai/models/*.toml are symlinks into providers/stepfun/models/ (the established stepfun-ai Global convention already on dev), so the Global step-3.7-flash resolves to the stepfun file, which declares [interleaved] field = "reasoning_content". The field is present via the symlink.
[medium] models/stepfun/step-3.7-flash.toml benchmark reconcile — done. The file now carries all 11 benchmarks: the 8 static.stepfun.com entries from dev plus the 3 OpenRouter-sourced ones, added alongside (none dropped).
[low] redundant [modalities] override on stepfun-ai — resolved. video lives on the base model (models/stepfun/step-3.7-flash.toml), and the Global entry (symlink) has no [modalities] block, so it inherits text/image/video.

Cited (PR body updated)

[low] step-1-32k / step-2-16k reasoning change — reasoning = false confirmed against StepFun's reasoning-models guide (zh / en), which lists only Step 3.7 Flash and Step 3.5 Flash (2603) as reasoning models. Step 1 / Step 2 are legacy chat models absent from that list; the prior reasoning = true on dev was incorrect.
[low] China per-token pricing — cited. StepFun pricing details (zh) list step-3.7-flash at ¥1.35/1M input (cache miss), ¥8.1/1M output, ¥0.27/1M input (cache hit); the TOML 0.185 / 1.11 / 0.037 are the USD conversions.
[low] video modality + knowledge cutoff — video is cited (same reasoning guide: "新增原生多模态输入能力，可直接理解图片和视频内容"). StepFun publishes no official knowledge cutoff, so 2026-03-01 stays an explicitly-disclosed estimate in the PR body.

Tianning Li added 2 commits May 30, 2026 01:37

Add Step 3.7 Flash model to stepfun and stepfun-ai providers

a19b200

Add step-3.7-flash.toml for both providers (released 2026-05-29, multimodal text/image/video input, 256K context, open weights) plus provider logos.

lit26 changed the title ~~Add Step 3.7 Flash model to stepfun and stepfun-ai providers~~ Add Step 3.7 Flash and StepFun Step Plan providers May 29, 2026

Fix StepFun China step-3.7-flash pricing and clarify provider names

467363a

- Convert step-3.7-flash China cost from copied global USD to RMB-converted USD (0.185/1.11/0.037 from ¥1.35/8.1/0.27), matching step-3.5-flash convention - Rename providers to disambiguate China vs Global variants

lit26 marked this pull request as ready for review June 1, 2026 12:43

Tianning Li and others added 4 commits June 6, 2026 16:40

Merge branch 'dev' into add-step-3.7-flash

c98b11d

Merge branch 'dev' into add-step-3.7-flash

2c779fc

Merge branch 'anomalyco:dev' into add-step-3.7-flash

054da7d

Merge branch 'anomalyco:dev' into add-step-3.7-flash

3763be6

Tianning Li added 2 commits June 15, 2026 00:08

Merge branch 'dev' into add-step-3.7-flash

8831574

lit26 added 4 commits June 15, 2026 23:33

Merge branch 'anomalyco:dev' into add-step-3.7-flash

b5bec57

Merge branch 'anomalyco:dev' into add-step-3.7-flash

c74a7d5

Merge branch 'anomalyco:dev' into add-step-3.7-flash

ca216d4

Merge branch 'anomalyco:dev' into add-step-3.7-flash

7463396

Tianning Li added 2 commits June 28, 2026 23:47

Merge branch 'dev' of https://github.com/anomalyco/models.dev into ad…

5f00eb5

…d-step-3.7-flash # Conflicts: # providers/stepfun-ai/provider.toml # providers/stepfun/models/step-3.7-flash.toml

fix(stepfun): restore China step-3.7-flash cost and add video modality

8eeea3e

rekram1-node reviewed Jun 28, 2026

View reviewed changes

Comment thread providers/stepfun-ai/provider.toml

Tianning Li added 2 commits June 29, 2026 16:57

fix(stepfun-ai): update reasoning HTTP format access date and clarify…

57e8b1a

… comments

Tianning Li added 2 commits July 1, 2026 00:32

Merge remote-tracking branch 'upstream/dev' into add-step-3.7-flash

996ad32

# Conflicts: # providers/stepfun/models/step-3.7-flash.toml

Merge remote-tracking branch 'upstream/dev' into add-step-3.7-flash

c3ede08

# Conflicts: # providers/stepfun-ai/models/step-3.7-flash.toml # providers/stepfun-ai/models/step-3.7-flash.toml~HEAD

lit26 requested a review from rekram1-node July 3, 2026 07:53

Merge branch 'anomalyco:dev' into add-step-3.7-flash

2cd05cd

rekram1-node reviewed Jul 7, 2026

View reviewed changes

Tianning Li added 2 commits July 9, 2026 17:10

Tianning Li and others added 2 commits July 9, 2026 17:48

Merge branch 'anomalyco:dev' into add-step-3.7-flash

4c419b3

lit26 requested a review from rekram1-node July 10, 2026 05:08

Tianning Li added 3 commits July 11, 2026 12:59

Merge remote-tracking branch 'upstream/dev' into add-step-3.7-flash

6111f36

# Conflicts: # models/stepfun/step-3.7-flash.toml

Merge branch 'add-step-3.7-flash' of https://github.com/lit26/models.dev

6cbc76a

into add-step-3.7-flash

fix(stepfun-ai-step-plan): use platform.stepfun.ai/en doc links

0c3e1b3

Global step-plan model docs now point to the .ai/en guides instead of the China .com/zh pages. Links verified live.

rekram1-node merged commit 12c9f29 into anomalyco:dev Jul 13, 2026
1 of 2 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Add Step 3.7 Flash and StepFun Step Plan providers#1903

Add Step 3.7 Flash and StepFun Step Plan providers#1903
rekram1-node merged 28 commits into
anomalyco:devfrom
lit26:add-step-3.7-flash

lit26 commented May 29, 2026 •

edited

Loading

Uh oh!

hengm3467 commented May 30, 2026

Uh oh!

lit26 commented May 30, 2026 •

edited

Loading

Uh oh!

rekram1-node commented Jun 13, 2026

Uh oh!

lit26 commented Jun 14, 2026

Uh oh!

javlund commented Jun 27, 2026

Uh oh!

Uh oh!

rekram1-node Jul 7, 2026

Uh oh!

lit26 Jul 9, 2026

Uh oh!

github-actions Bot commented Jul 9, 2026

Uh oh!

github-actions Bot commented Jul 9, 2026

Uh oh!

github-actions Bot commented Jul 11, 2026

Uh oh!

lit26 commented Jul 11, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Uh oh!

Conversation

lit26 commented May 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Step Plan providers

Step 3.7 Flash details

Sources / citations

Test plan

Uh oh!

hengm3467 commented May 30, 2026

Uh oh!

lit26 commented May 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rekram1-node commented Jun 13, 2026

Uh oh!

lit26 commented Jun 14, 2026

Uh oh!

javlund commented Jun 27, 2026

Uh oh!

Uh oh!

rekram1-node Jul 7, 2026

Choose a reason for hiding this comment

Uh oh!

lit26 Jul 9, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Jul 9, 2026

Review: PR #1903 — Add Step 3.7 Flash and StepFun Step Plan providers

Hard blockers: None ✅

Non-blocking recommendations

Summary

Uh oh!

github-actions Bot commented Jul 9, 2026

Uh oh!

github-actions Bot commented Jul 11, 2026

Action items

Uh oh!

lit26 commented Jul 11, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

lit26 commented May 29, 2026 •

edited

Loading

lit26 commented May 30, 2026 •

edited

Loading