CostPerPrompt Pricing 2026

Agent-native notebook that finds data, writes cells, provisions GPUs and runs experiment branches from a prompt

Pricing: Usage-based with four published tiers plus Enterprise. Free is $0/mo: the Auto model for fast low-cost everyday work on a CPU sandbox with 8 vCPU and 8 GB RAM. Plus is $30/mo, adding open models (DeepSeek V4 Pro, Kimi K3, Qwen3.8 Max), entry GPUs (T4, L4, A10) up to 24 GB VRAM, up to 32 GB sandbox RAM, a monthly usage allowance and optional pay-as-you-go beyond it. Pro is $90/mo for every model including Claude Sonnet and Opus and GPT, mid-tier GPUs (L40S, A100) up to 80 GB VRAM, up to 64 GB sandbox RAM and at least 4× the Plus allowance. Max is $200/mo with all models and GPUs unthrottled, H100 and H200 up to 141 GB VRAM, up to 128 GB sandbox RAM, at least 12× the Plus allowance and priority capacity. Enterprise covers higher limits, dedicated capacity and SSO.

Chat with your data for instant analysis

Pricing: Free tier. Pro $20/mo with advanced features

Japanese equities API where every figure links to its source filing, with 60-names-to-one entity resolution

Pricing: Three tiers, all reading the same verified source-linked filings — the company states explicitly that you do not pay for better data, only for more surfaces and higher limits, and that tier upgrades take effect within seconds and are enforced at the API. Free is $0 per month with 1,000 requests per day at 2 per second, covering annual financial reports, company search, ratios and sectors, and the filing list and calendar, plus the typed Python SDK and OpenBB Workspace integration. Developer, marked recommended, is $199 per month with 10,000 requests per day at 5 per second and adds semi-annual reports, filing sections with translations, time-series with growth and CAGR, screening, rankings and compare, health scores and buybacks, shareholdings, board composition and AGM voting, earnings actuals and forecasts, the MCP server with 40+ AI tools, webhook filing alerts and 24-hour email support. Institutional is contact-priced and adds quarterly reports, trajectories and accumulation signals, the activist tracker and unwinding scoreboards, the capital-allocation classifier, the per-row purpose audit bundle, the relationship graph, bulk Parquet export, custom rate limits, dedicated same-day support, a 99.9% uptime SLA and custom redistribution rights. Cancel anytime.

AI forecasting engine with a public Metaculus and real-money track record, priced per question

Pricing: Free tier starts with $20 in credit, no card, 10 concurrent researchers, all operations and API access. Analyst Starter is $20/month with 20 concurrent researchers and 15% off usage rates; Research is $99/month (half off the first month for new subscribers) with 40 researchers and 30% off usage; Expert is $500/month with 80 researchers and 30% off usage. Usage is charged per question on top: roughly $2–$5 for a high-effort forecast, $0.25–$3 for a low-effort one, $0.30–$4 per multi-agent row and $0.03–$0.30 per agent-map row. Top-up credit can be bought in any amount, is good for a year and does not reset monthly like the plan allowance. Enterprise and volume pricing on request.

Open-source agentic GenBI with a governed semantic layer, serving both humans and AI agents across 20+ sources

Pricing: Free is $0/month with 20 free credits per month, 80 credits for the first 14 days, Agentic and Classic modes, knowledge base and visualisation, 2 projects, 2 members and 10 tables per project. Essential Cloud is $179/month billed annually with 13,200 annual credits included. Enterprise Cloud is $559/month billed annually with 24,000 annual credits and adds row-level security and further data-security controls. Cloud plans are usage-based: credits roll over up to 2× and additional credits are $0.10 each. The same plan set covers cloud and self-hosted deployments. Monthly free credits expire at month end and do not roll over.