OneInterface routes coding and chat work to high cost-performance models commonly preferred for that kind of task, with transparent cost and fallback behavior.
OneInterface Routing Profiles
Experience-based routing layer, not a black-box router
Task Contexts
Users stay in their existing tools and choose the task context. The routing profile reflects what builders typically prefer for that kind of work.
Coding
oi/cheap-coderFor teams that usually pick the lowest-cost capable model for small edits, tests, comments, and background agent work.
Common choice: deepseek-v4-flash
Baseline: claude-sonnet-4
oi/balanced-coderFor everyday coding tasks where builders usually want a practical balance of quality, speed, and cost.
Common choice: qwen3.6-plus
Baseline: claude-sonnet-4
oi/deep-reviewFor review and bug-hunting work where users tend to prefer more deliberate reasoning.
Common choice: deepseek-v4-pro
Baseline: claude-opus-4
oi/long-contextFor repository analysis, large diffs, logs, and multi-file planning where long context matters most.
Common choice: kimi-k2.6
Baseline: claude-sonnet-4
Chat
oi/chatFor everyday assistant and conversational workloads where builders want a practical balance of quality, speed, and cost.
Common choice: kimi-k2.6
Baseline: claude-sonnet-4
oi/chat-fastFor high-volume, latency-sensitive chat where the lowest-cost capable model is preferred.
Common choice: deepseek-v4-flash
Baseline: claude-sonnet-4
oi/chat-smartFor chat that needs more deliberate reasoning — complex questions, analysis, and nuanced writing.
Common choice: deepseek-v4-pro
Baseline: claude-opus-4
Agent
oi/agent-fastFor high-volume, latency-sensitive agent loops where the cheapest capable tool-calling model is preferred.
Common choice: qwen3.6-plus
Baseline: claude-sonnet-4
oi/agentFor everyday agentic work where builders want a practical balance of tool-calling quality, speed, and cost.
Common choice: kimi-k2.6
Baseline: claude-sonnet-4
oi/agent-maxFor agent work that needs more deliberate reasoning — complex multi-step planning and hard tool orchestration.
Common choice: deepseek-v4-pro
Baseline: claude-opus-4
Why This Wedge
Model-centric products ask developers to compare dozens of IDs. Experience-based routing starts from the job to be done: small edits, daily coding, deep review, or long-context work.
Transparent routing
Usage rows show requested routing profile, actual model, tokens, cost, and estimated savings.
High cost-performance supply
Routing profiles reflect practical model preferences for each coding workflow.
Compatibility first
Works through OpenAI-compatible APIs and the Anthropic Messages proxy for Claude Code.
Use oi/balanced-coder for everyday coding work, then switch only when the task needs lower cost, deeper review, or longer context.