本文へ移動
メイン ドキュメント
ドキュメント

Models

The catalog at GET /v1/models is the truth. Models belong to one of two teams; the team decides the route and the admission bar. Pass the id as model unchanged.

Shell 
curl -s https://api.kotoba.cloud/v1/models -H "Authorization: Bearer $KOTOBA_API_TOKEN"
# → data[].id · team · route · availability

Teams and admission

Current catalog

A model whose route is not configured is refused by name at admission and spends no quota. Which providers stand behind each route is operator configuration and is not stated here.

Per-model detail

qwen/qwen3.8-flash

A general model on the shared route. Sign-in only, no verification; the three standard tasks (code-review / vulnerability-triage / remediation). The id is the publisher's, used unchanged.

  • Team: Blue team · route: shared · served
  • Admission: a signed-in principal, free quota, guardrails (the three standard tasks only)
  • Ceilings: 524,288 input characters · 32,768 output tokens (reasoning included) · 128 tools · 256 turns
  • The answer's model echoes qwen/qwen3.8-flash (strictly matched)
Shell 
curl -s -X POST https://api.kotoba.cloud/v1/chat/completions -H "Authorization: Bearer $KOTOBA_API_TOKEN" \
  -H "content-type: application/json" -d '{"model":"qwen/qwen3.8-flash","max_tokens":2048,"messages":[{"role":"user","content":"..."}]}'

z-ai/glm-5.3-flash

The other general model on the shared route; the same admission and ceilings as qwen/qwen3.8-flash.

  • Team: Blue team · route: shared · served
  • Admission: a signed-in principal, free quota, guardrails (the three standard tasks only)
  • Ceilings: 524,288 input characters · 32,768 output tokens (reasoning included) · 128 tools · 256 turns
  • The answer's model echoes z-ai/glm-5.3-flash (strictly matched)
Shell 
curl -s -X POST https://api.kotoba.cloud/v1/chat/completions -H "Authorization: Bearer $KOTOBA_API_TOKEN" \
  -H "content-type: application/json" -d '{"model":"z-ai/glm-5.3-flash","max_tokens":2048,"messages":[{"role":"user","content":"..."}]}'

qwen3.8-flash-next-whitehacker

A flash-class model tuned for security research: code review, vulnerability triage, remediation. It reasons before answering and returns native tool calls. Unrestricted for authorized researchers; AUP consent is the premise.

  • Team: Red team · route: dedicated · served
  • Admission: card verification, AUP consent, approved scope (code-review comes with verification), free quota, guardrails
  • Ceilings: 524,288 input characters · 32,768 output tokens (reasoning included) · 128 tools · 256 turns
  • The answer's model echoes qwen3.8-flash-next-cybersecurity-nvfp4 (strictly matched)
Shell 
curl -s -X POST https://api.kotoba.cloud/v1/chat/completions -H "Authorization: Bearer $KOTOBA_API_TOKEN" \
  -H "content-type: application/json" -d '{"model":"qwen3.8-flash-next-whitehacker","max_tokens":2048,"messages":[{"role":"user","content":"..."}]}'

glm5.3-flash

The red team's second model. Its dedicated origin is not up yet, so it is listed and refused by name at admission (no quota spent). This page flips to “served” when the origin stands.

  • Team: Red team · route: dedicated · route not configured
  • Admission: card verification, AUP consent, approved scope (code-review comes with verification), free quota, guardrails
  • Ceilings: 524,288 input characters · 32,768 output tokens (reasoning included) · 128 tools · 256 turns
  • The answer's model echoes glm5.3-flash (validated on shape and content)
Shell 
curl -s -X POST https://api.kotoba.cloud/v1/chat/completions -H "Authorization: Bearer $KOTOBA_API_TOKEN" \
  -H "content-type: application/json" -d '{"model":"glm5.3-flash","max_tokens":2048,"messages":[{"role":"user","content":"..."}]}'