Docs | Second Silicon | Docs

Second Silicon runs each thread on a model you choose, or picks one for you from the providers you’ve configured. You can pin a model from the composer, or bring your own API key to use your existing provider quota.

Supported providers

Provider	Models	Best for
Anthropic	Claude Opus 4.8, Claude Sonnet 4.6	Deep reasoning and multi-step debug
OpenAI	GPT-5.5, GPT-5.4, GPT-5.4 mini	General reasoning and faster, lower-cost runs
OpenRouter	Kimi-k2.6, GLM 5.2	Alternative routing through OpenRouter
Google	Gemini 2.5 Pro, Gemini 2.5 Flash	Multimodal and fast general-purpose runs

The model picker beside the chat composer switches the model for the current thread. Settings → Models controls which models appear in that picker — disable the ones you don’t want to use.

Automatic selection

When you don’t pin a model, Second Silicon picks one from the providers you have a key for, in a cost-aware preference order. Pin a model from the composer whenever you want a specific one for the thread.

A separate per-thread reasoning effort setting controls how hard the model works before answering. Higher effort is more thorough but slower and more expensive.

Bring your own key (BYOK)

Add a provider API key to use your existing credits and keep full control over spend.

Open Settings

Click the gear icon at the bottom of the sidebar, then go to API Keys.

Paste your key

Paste the key into the Anthropic, OpenAI, OpenRouter, or Google field. Keys save automatically and are stored encrypted in your workspace.

Pick a model

Enable the provider’s models under Settings → Models, then pick a model from the composer or let Second Silicon choose.

If no BYOK key is set for a provider, availability depends on the shared keys enabled for your workspace, and usage limits may apply during high-demand periods.

Token usage and cost

Every message in a debug thread shows three numbers: input tokens, output tokens, and an estimated USD cost. Input tokens are the context sent to the model. Output tokens are the model’s response.

Cost is an estimate based on the selected provider, model family, and current pricing assumptions in Second Silicon. It can differ from the provider’s bill, especially when a provider changes prices, applies discounts, or routes through a different underlying model.

Cost basis

Usage views label the pricing basis used for each estimate, such as Claude Sonnet (est.), GPT-5.4 mini (est.), Gemini Flash (est.), Kimi K2.6 via OpenRouter (est.), or GLM 5.2 via OpenRouter (est.). If a specific model has no exact match, Second Silicon falls back to the closest provider-level estimate.

Verify current pricing on the provider’s site before making cost decisions.