Claude Sonnet 5.5 is the best model for Oh My Pi's default role: it scores 61.8% on Terminal-Bench 4.0 at half the price of Claude Opus 5.5. omp routes work by role (default, smol, slow, plan, advisor and more), so the useful answer is one model per role: Sonnet 5.5 for default, Opus 5.5 (64.9%) for slow and plan, Claude Haiku 5.5 for smol subagent fan-out, and a model from a second vendor, such as GPT-6.1 Sol, for the advisor role that reviews each turn.
Picked a model? One key can route each step to the cheapest capable one. See plans and prices →
Running Oh My Pi across a team? An LLM API for your team, with a key and a fixed budget for every member →
Sources: oh-my-pi README · Terminal-Bench 4.0 leaderboard · Anthropic on Haiku 5.5 · DeepSeek V4.1 Flash release
Run any of these models in your agent
Or let the router send each step to the cheapest model that can do it, on providers we vet for price, uptime and speed. Plans from $19 a month.
Best overall (default role). 61.8% on Terminal-Bench 4.0 with strong tool use, which omp's 31 built-in tools, LSP and debugger integration lean on.
Best for long agent runs (slow and plan roles). #1 on Terminal-Bench 4.0 at 64.9%. Its doubled price matters less in the roles that run less often.
Best value (smol role). Anthropic reports 39.2% on Terminal-Bench 4.0 at $0.10/$0.50 per million tokens, a good fit for cheap subagent fan-out with short prompts.
Advisor role. 58.2% on Terminal-Bench 4.0. A reviewer from a different vendor than the main model is more likely to catch what it missed.
Budget pick. Open weights at $0.30/$1.20 per million tokens at peak hours and half off-peak, for the commit and tiny roles; DeepSeek reports 42.3% on Terminal-Bench 4.0 for its October 17 update, up from 31.2%.
Open-weight pick. The top open-weight model on Terminal-Bench 4.0 (41.8%). Z.AI's GLM Coding Plan is one of omp's built-in providers.
The agent matters as much as the model. Compare AI agents →
Every model on this page runs on Standard Compute