Hacker News
CostPerPrompt Tracks Live Pricing for 232 Models With Workload Calculators for Agents, RAG and Voice
This Show HN aggregates per-million-token input, output and cached pricing across 232 models from 14 vendors including OpenAI, Anthropic, Google, DeepSeek, Alibaba, Moonshot, xAI and Z.ai, then layers six workload calculators on top — general comparison, chatbot with conversation history, agent workflows that account for tool use and retries, RAG split across indexing/retrieval/generation, voice bundling STT+LLM+TTS, and GPU rental. Current listings include GPT-5.6 Sol at $5/$30, Claude Opus 5 at $5/$25, Claude Sonnet 5 at $2/$10 and DeepSeek V4 Flash at $0.09/$0.18. The agent calculator is the differentiator; most pricing tables ignore retry and tool-call amplification, which is where real agent bills come from.
↳ Follow the thread