Baseten Integration

Kimi K3 via Baseten, explicitly routed

PorkiCoder names every Baseten-backed path so users know where paid work runs. Use the distinct Kimi K3 consult for a Baseten-hosted second opinion, or use the existing managed fan-out and HogCode routes for their supported Kimi and GLM workflows.

Download PorkiCoder Read Consult MCP docs

Current Availability

Available

Kimi K3 via Baseten

consult_kimi3_baseten is a distinct, explicitly approved, metered second-opinion route through PorkiCoder's authenticated proxy.

Available

Managed fan-out and HogCode

Kimi K2.7 Code and GLM-5.2 can run as isolated fan-out candidates. GLM-5.2 can judge results, while HogCode uses its managed Kimi K2.7 Code route.

Retired

Legacy Consult names

consult_kimi, consult_kimi27, consult_kimi27_fast, and consult_glm52 remain retired. They are not aliases for the new Kimi K3 route.

Use the Available Routes

The Baseten-backed paths are wired into PorkiCoder's managed workflows. Choose the named route you intend to authorize; PorkiCoder never substitutes one provider for another.

1
Install PorkiCoder
Download PorkiCoder, sign in, and open a project.
2
Authorize the route
Explicitly approve Kimi K3 via Baseten, or approve the fan-out model mix, worker count, and budget for this conversation.
3
Start the workflow
Call consult_kimi3_baseten, run Kimi K2.7 Code or GLM-5.2 in isolated worktrees, or launch hogcode.
4
Review the result
Inspect candidate diffs before applying anything. Judging and apply operations remain separate, explicit steps.

Baseten-Backed Models

Each provider path is named and scoped so users can tell where a paid request runs.

Route Transport Purpose
Kimi K2.7 Code PorkiCoder-managed Baseten routing for fan-out and HogCode. Isolated implementation or review candidates; read-only project assistance in HogCode.
GLM-5.2 PorkiCoder-managed Baseten routing for fan-out workers and judges. Large-context candidate reviews, architecture tradeoffs, and fan-out judging.
Kimi K3 via Baseten
consult_kimi3_baseten		 Authenticated PorkiCoder proxy; the production Baseten credential remains server-side. An explicitly approved, metered Kimi K3 second opinion with clear Baseten attribution.

Routing, Consent, and Credentials

  • PorkiCoder keeps provider routes distinct. consult_kimi3 uses Kimi Platform; consult_kimi3_baseten uses Baseten. Neither silently fails over to the other.
  • Paid fan-out, judging, and consult calls require explicit current-conversation approval. Consent is not inferred from an earlier request.
  • PorkiCoder handles authenticated routing, streaming, quota enforcement, usage tracking, and bounded transient-failure handling on managed routes.
  • The production Baseten credential stays server-side. PorkiCoder does not inject it into terminal environments or the Consult MCP process.
No direct Baseten BYOK fallback
The new route is designed to be proxy-only. Creating extra API keys does not create documented extra workspace capacity, and PorkiCoder will not use terminal-injected Baseten keys as a quota workaround.

Kimi K3 Capacity Gate

As checked on August 10, 2026, Baseten publishes default Model API limits of 120 requests per minute and 1,000,000 tokens per minute for Pro. Verified Basic and Pro publish the same default 120 RPM, so dependable scale still requires workload-appropriate sustained and burst capacity. PorkiCoder queues shared work and handles explicit rate-limit failures, but does not multiply quotas with extra keys.

Baseten's published Kimi K3 Model API rates are $3.00 per million uncached input tokens, $0.30 per million cached input tokens, and $15.00 per million output tokens. Provider pricing and limits can change; the official pages remain the source of truth.

Example Prompts

Name the route and authorization in plain English:

  • Run a two-worker fan-out with Kimi K2.7 Code and GLM-5.2, then compare the diffs.
  • Use GLM-5.2 as the fan-out judge and explain the architectural risks in the winning diff.
  • Launch HogCode and explain this project using its read-only tools.
  • I authorize one paid Kimi K3 via Baseten call. Use consult_kimi3_baseten to review this architecture.

FAQ

Do users need a Baseten account or API key?
Not for PorkiCoder-managed routes. PorkiCoder keeps its production credential server-side and applies its own authentication, consent, quota, and usage controls.
What happens when shared capacity is exhausted?
PorkiCoder returns a clear rate-limit or capacity error. It does not retry at both app and backend layers, switch providers silently, or use additional API keys as a quota workaround.
Where can I read the general docs?
See the worktree fan-out section, the HogCode section, and the separate Consult MCP section for the current consult tool catalog.