| # | Name | Creator | Released | In $/M | Out $/M | Blended | Intelligence | Coding | Agentic | GPQA % | HLE % | IFBench % | ΟΒ²-Bench % | LCR % | GDPval % | SciCode % | TerminalBench % | Omniscience % | Non-Halluc % | Context | Value/$ | Safety | Spd | Lat | Frontier | Slug |
|---|
| # | Name | Creator | Released | In $/M | Out $/M | Blended | Intelligence | Coding | Agentic | GPQA % | HLE % | IFBench % | ΟΒ²-Bench % | LCR % | GDPval % | SciCode % | TerminalBench % | Omniscience % | Non-Halluc % | Context | Value/$ | Safety | Spd | Lat | Frontier | Slug |
|---|
Ranked by fit. Premium = OpenAI/Anthropic/Google, used directly via their own API. Standard/Budget below are available via OpenRouter.
Architecture, hard debugging, anything worth paying premium for
Complex refactors, unfamiliar codebases, architecture-level code review
Check-ins, briefings, reminders, conversational coaching
Debugging, scripts, codebase navigation, agent workflows
Check-ins, briefings, reminders, conversational coaching
MiMo-V2.5 β $0.14/$0.28 per M (free-tier cost equivalent) Β· (paid-tier per-task pending) Β· cheapest safe option for unattended coaching
Debugging, scripts, codebase navigation, agent workflows
Questions that don't need tools, files, or automation β use the free web interface directly and skip API costs.