Open-weight LLM landscape

ZZhipu AIAAlibabaMMoonshot AIDDeepSeekMMiniMax (minimax.io)TThinking MachinesTTencentXXiaomiNNvidiaMMetaSStepFun (China)MMistralGGoogleOOpenAICCohereMMistralAArcee AI

Blended cost = (7·cache + 2·input + 1·output)/10. When a provider omits cache-hit pricing, the input price is used as an upper bound (cache hits are never more expensive than a regular input token). 44 models shown: open-weight (via models.dev), with an Agentic Index score AND input + output pricing. Models without input or output pricing are excluded. Neuralwatt energy values in tooltips are measured when an AA model matches a NW model; otherwise the tooltip shows an estimated energy derived from the NW cost ↔ energy regression (see the Neuralwatt scatter below).

X: NW blended cost per 1M tokens = (7·cache + 2·input + 1·output)/10 (USD, from Neuralwatt pricing). Y: energy per request at the 16k–64k prompt-size band (mWh, scraped from the Neuralwatt portal). Dashed line: linear regression on base models only (excluding -fast / -short variants). Hover any dot for details.