fix: mark Kimi-K3 non-reasoning to match supports_reasoning = false by wynandhuizinga · Pull Request #2 · NEBUL-AI/models.dev

wynandhuizinga · 2026-09-23T07:43:56Z

Resolves the high-severity check on providers/nebul/models/moonshotai/Kimi-K3.toml in anomalyco#7400.

GET /v1/model/info (re-checked 2026-09-22) marks this served ID supports_reasoning = false and exposes no reasoning_efforts; live completion returns content only, no reasoning_content trace. (run the curl before merging to confirm the second half)
The entry now overrides the inherited lab default with reasoning = false and drops reasoning_options / [interleaved].
Sync module: authored reasoning = false is now the documented escape hatch — without it, hourly sync would throw MissingReasoningOptionsError and re-add the contradictory reasoning_options = [].
Verified: bun validate passes, 18/18 Nebul sync tests pass, dry-run sync round-trips the file unchanged.

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

…alyco#7805)

Live on https://kenari.id/v1/models as deepseek-v4-1-flash (context 1M, input text+image, reasoning low/high/max, tool_call). Uses base_model deepseek/deepseek-v4.1-flash, cost 0/0 per Kenari subscription convention. Co-authored-by: davidstrouk <david@davidstrouk.com>

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

) * fix(xiaomi): mark MiMo V2.6 Flash/Pro open-weight and drop unsupported pdf input UltraSpeed remains closed-weight. Align input modalities to text/image/audio/video. * docs(xiaomi): link MiMo V2.6 Flash/Pro Hugging Face weights * fix(xiaomi): mark MiMo V2.6 Pro-UltraSpeed open-weight with Pro-RL weights Same parameters as Pro; link the shared Hugging Face release. * fix(xiaomi): drop incorrect MiMo V2.6 knowledge cutoff The 2026-09-22 value was the release date, not a knowledge cutoff. Omit until a real cutoff is known.

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

…3-contributor (anomalyco#7729) * feat(cline-pass): add mimo-v2.6 and muse-spark-1.3 models Live clinePass catalog lists 14 models; three slugs were absent from providers/cline-pass/models/cline-pass/. All three serve requests on api.cline.bot (probe-verified 2026-09-22); nonexistent slugs 404. muse-spark-1.3-contributor carries Meta reference rates because Cline publishes no ClinePass rate for it. * fix(cline-pass): follow review — wire-cited effort, drop muse cost mimo-v2.6-* reasoning_options now list exactly the seven levels the ClinePass endpoint accepts (probe 2026-09-22: seven levels -> 200, unknown level -> 500), cited in a leading wire comment. muse-spark-1.3-contributor drops [cost]: Cline publishes no ClinePass rate for the slug. * fix(cline-pass): revert mimo effort to same-host peer set Effect probe (reasoning_tokens, 3 trials/level) shows minimal/max do not differentiate on this host, so declare the mimo-v2.5* peer set per review; wire comment keeps the full acceptance evidence.

Adds the Step 5 Preview lab entry and serves it from all four StepFun hosts: StepFun (China), StepFun (Global), and both Step Plan channels. Specs from the official model page: 1M context / input / output, text + image + video input, text output, tool calling, JSON Mode + JSON Schema, prompt caching, and low/medium/high reasoning effort via Chat `reasoning_effort`, Messages `output_config.effort`, and Responses `reasoning.effort`. Pricing: China converts the CNY list price (7 / 20 / 0.35 per 1M) at the same 7.2973 CNY/USD rate the other StepFun entries use; Global uses the published USD price (1.00 / 2.70 / 0.05) directly, so that entry is a real file rather than a symlink into providers/stepfun. Step Plan entries carry no cost block, matching their siblings — that channel bills against plan credit, not per token. Provider comments also updated: step-5-preview added to the effort matrix, and the Responses API now serves step-5-preview alongside step-3.7-flash. Sources: https://platform.stepfun.com/docs/zh/guides/models/step-5-preview https://platform.stepfun.ai/docs/en/guides/models/step-5-preview https://platform.stepfun.com/docs/zh/guides/pricing/details https://platform.stepfun.ai/docs/en/guides/pricing/details https://platform.stepfun.com/docs/zh/step-plan/integrations/reasoning-api Co-authored-by: Tianning Li <litianning@stepfun.com>

Co-authored-by: Aiden Cline <rekram1-node@users.noreply.github.com>

…try shows none/low/medium/high/xhigh, but DeepSeek V4.1 Flash officially supports off/low/high/max (anomalyco#7534) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

…false

…easoning

opencode-agent Bot and others added 30 commits September 20, 2026 22:24

chore(sync): update OpenRouter model catalog (anomalyco#7598)

28edbfd

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update OpenRouter model catalog (anomalyco#7600)

27efbe9

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Kilo model catalog (anomalyco#7601)

9c20af0

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Kilo model catalog (anomalyco#7603)

91c2e78

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update OpenRouter model catalog (anomalyco#7604)

c44d7ad

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update NanoGPT model catalog (anomalyco#7602)

34e14cb

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update OpenRouter model catalog (anomalyco#7606)

fdee770

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Kilo model catalog (anomalyco#7605)

76488ff

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Kilo model catalog (anomalyco#7608)

62c3e41

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update OpenRouter model catalog (anomalyco#7609)

52eff77

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update OpenRouter model catalog (anomalyco#7611)

b44a343

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Kilo model catalog (anomalyco#7610)

3764983

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update NanoGPT model catalog (anomalyco#7613)

1c43b10

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Kilo model catalog (anomalyco#7612)

4e14674

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update OpenRouter model catalog (anomalyco#7614)

2a91d70

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Eden AI model catalog (anomalyco#7616)

cfc5669

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update OpenRouter model catalog (anomalyco#7618)

5ec54bd

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update NanoGPT model catalog (anomalyco#7621)

b8647dc

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update LLM Gateway model catalog (anomalyco#7619)

5cd0647

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Kilo model catalog (anomalyco#7620)

d8d98a4

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update NanoGPT model catalog (anomalyco#7623)

962a7ce

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Cortecs model catalog (anomalyco#7624)

25c48bf

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update NanoGPT model catalog (anomalyco#7628)

60efdd3

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update OpenRouter model catalog (anomalyco#7629)

5f15e38

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Kilo model catalog (anomalyco#7627)

b3c1811

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update NanoGPT model catalog (anomalyco#7630)

1c56cae

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Kilo model catalog (anomalyco#7632)

48e0aaa

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update OpenRouter model catalog (anomalyco#7631)

cfcc239

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update OpenRouter model catalog (anomalyco#7634)

e379bc9

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Kilo model catalog (anomalyco#7633)

074421e

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

github-actions Bot and others added 29 commits September 22, 2026 22:57

fix: [missing-model] ofox: x-ai/grok-4.7 (anomalyco#7792)

c3fd2c8

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

fix: [missing-model] pioneer: fastino/gliguard-PII-multi (anomalyco#7780

3960b9b

) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

fix: [missing-model] ofox: openai/gpt-6-luna (anomalyco#7804)

7d4d667

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

feat(amazon-bedrock): add GPT-6 Sol and Luna inference profiles (anom…

d53018c

…alyco#7805)

fix: [missing-model] github-copilot: claude-opus-5.5 (anomalyco#7815)

1ec5c57

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

fix: [missing-model] ofox: openai/gpt-6-sol (anomalyco#7803)

80f9e2b

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

chore(sync): update DigitalOcean model catalog (anomalyco#7789)

3487ff2

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Cloudflare Workers AI model catalog (anomalyco#7707)

9755563

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

fix: add supported Kimi K3 Bedrock profiles (anomalyco#7795)

1222d43

Co-authored-by: Aiden Cline <rekram1-node@users.noreply.github.com>

fix: ClinePass thinking-effort levels are wrong for all models: regis…

07fb48a

…try shows none/low/medium/high/xhigh, but DeepSeek V4.1 Flash officially supports off/low/high/max (anomalyco#7534) Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

chore(sync): update NanoGPT model catalog (anomalyco#7839)

76689e2

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update OpenRouter model catalog (anomalyco#7840)

7d4b261

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update EmpirioLabs AI model catalog (anomalyco#7837)

e4eb305

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Kilo model catalog (anomalyco#7838)

59eda21

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update CrossModel model catalog (anomalyco#7841)

7d87c4f

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Kilo model catalog (anomalyco#7845)

1c94dc6

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Eden AI model catalog (anomalyco#7847)

b3318a6

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update OpenRouter model catalog (anomalyco#7846)

d149193

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update NanoGPT model catalog (anomalyco#7848)

f380ec9

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update OpenRouter model catalog (anomalyco#7849)

ceec55e

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Kilo model catalog (anomalyco#7850)

812f7a1

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update OpenRouter model catalog (anomalyco#7855)

519f233

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

chore(sync): update Kilo model catalog (anomalyco#7854)

e346898

Co-authored-by: opencode-agent[bot] <opencode-agent[bot]@users.noreply.github.com>

fix(nebul): mark Kimi-K3 non-reasoning to match supports_reasoning = …

b02a765

…false

Merge remote-tracking branch 'origin/dev' into wh/fix/nebul-kimi-k3-r…

30f8879

…easoning

wynandhuizinga merged commit 9e67ffa into dev Sep 23, 2026
2 checks passed

Navigation Menu

fix: mark Kimi-K3 non-reasoning to match supports_reasoning = false - #2

fix: mark Kimi-K3 non-reasoning to match supports_reasoning = false#2
wynandhuizinga merged 398 commits into
devfrom
wh/fix/nebul-kimi-k3-reasoning

wynandhuizinga commented Sep 23, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

20 participants

Conversation

wynandhuizinga commented Sep 23, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

20 participants