{"generated_at":"2026-06-05T22:11:48.576Z","indices":{"TCI-GPU":{"value":5.2639,"value_prior":5.3132,"delta_pct":-0.9290474969357356,"constituents":[{"key":"h100-sxm5-80gb","name":"H100 SXM5","value":5.6148},{"key":"h200-sxm5-141gb","name":"H200 SXM5","value":4.5507},{"key":"b200-sxm-180gb","name":"B200 SXM","value":6.89},{"key":"a100-sxm4-80gb","name":"A100 SXM4","value":4}],"last_observed":"2026-06-05 21:42:10.124551+00"},"TCI-Inf-Open":{"value":0.922,"value_prior":0.922,"delta_pct":0,"constituents":[{"key":"deepseek/deepseek-chat-v3.1","name":"DeepSeek: DeepSeek V3.1","value":0.7899999999999999},{"key":"deepseek/deepseek-r1","name":"DeepSeek: R1","value":2.5},{"key":"meta-llama/llama-3.3-70b-instruct","name":"Meta: Llama 3.3 70B Instruct","value":0.32},{"key":"meta-llama/llama-4-maverick","name":"Meta: Llama 4 Maverick","value":0.6},{"key":"qwen/qwen-2.5-72b-instruct","name":"Qwen2.5 72B Instruct","value":0.39999999999999997}],"last_observed":"2026-06-05 21:41:42.5611+00"},"TCI-Inf-Closed":{"value":9.167,"value_prior":9.167,"delta_pct":0,"constituents":[{"key":"anthropic/claude-sonnet-4.6","name":"Anthropic: Claude Sonnet 4.6","value":15},{"key":"google/gemini-2.5-flash","name":"Google: Gemini 2.5 Flash","value":2.5},{"key":"openai/gpt-4o","name":"OpenAI: GPT-4o","value":10}],"last_observed":"2026-06-05 21:41:42.5611+00"}},"methodology":{"version":"2026-04-20","indices":{"GPU-Hour Index":{"description":"On-demand $/GPU-hr across four standardized GPU configurations, taken cross-provider and rolled up into a single number.","constituents":["h100-sxm5-80gb","h200-sxm5-141gb","b200-sxm-180gb","a100-sxm4-80gb"],"unit":"USD per GPU-hour","calculation":"For each constituent SKU we take the cross-provider median of $/GPU-hr (price_per_hour_usd ÷ gpu_count) over the last 24 hours. The index is the arithmetic mean of those four SKU medians.","sources":["dstackai/gpuhunt (hyperscalers + neoclouds)","Vast.ai bundles API (marketplace listings)"],"refresh":"hourly"},"Token Index · Open Frontier":{"description":"Output token price across five open-weights frontier models, averaged into a single $/Mtoken figure.","constituents":["meta-llama/llama-3.3-70b-instruct","meta-llama/llama-4-maverick","deepseek/deepseek-r1","deepseek/deepseek-chat-v3.1","qwen/qwen-2.5-72b-instruct"],"unit":"USD per million output tokens","calculation":"For each constituent we take the latest OpenRouter-published output price, scale per-token to per-million (×1e6), and average across constituents.","sources":["OpenRouter /api/v1/models"],"refresh":"hourly"},"Token Index · Closed Frontier":{"description":"Output token price across three closed frontier models, averaged into a single $/Mtoken figure.","constituents":["anthropic/claude-sonnet-4.6","openai/gpt-4o","google/gemini-2.5-flash"],"unit":"USD per million output tokens","calculation":"Same calculation as the Open index, applied to closed-frontier constituents.","sources":["OpenRouter /api/v1/models"],"refresh":"hourly"}},"standardized_unit_spec":{"gpu":{"vendor":"NVIDIA","example":"H100 SXM5 80GB with NVLink interconnect, 8-GPU node","pricing_type":"on-demand","excluded":"spot, reserved, committed-use discounts"},"inference":{"example":"1 million output tokens at the published on-demand rate per constituent model","excluded":"batched/cached pricing, free-tier promotional rates"}},"notes":"Values are point-in-time snapshots built from public provider catalogs. They are not transaction-weighted and are not audited to settlement-grade standards."}}