An analysis of 100T+ tokens from the past year shows reasoning models now represent over half of all usage, open-weight model use has grown steadily, and more
OpenRouter’s dataset offers a demand-side view of model selection across a large, multi-provider routing platform: reasoning-oriented systems have become the dominant usage category while open-weight options continue to gain share.
The finding is an early marker of a market that later showed lower-cost Chinese models overtaking US rivals in token consumption and sustained US-company demand for Chinese models on the same routing platform. That makes routing data consequential not just as a usage metric, but as evidence of where buyers are finding usable performance and price.
First-order effects
Reasoning-model providers gain a clear demand signal: more than half of observed usage is now concentrated in workloads that justify inference-time reasoning rather than simple generation.
Open-weight model developers and their hosting partners gain validation that availability through a router can translate into steadily expanding real-world use.
Second-order effects
Model buyers have stronger incentive to route requests by task, trading off reasoning capability, cost, and model openness rather than standardizing on a single provider.
Closed-model vendors face pressure from the combination of reasoning demand and growing open-weight adoption; later usage data showing lower-cost Chinese models gaining token share sharpens that competitive comparison.
Third-order effects
If these usage patterns persist, the durable advantage may accrue increasingly to routing and inference layers that let customers switch among models, rather than solely to any one model maker.
The same flexibility can make model-access policy more economically consequential: later reporting that Chinese models reached substantial US-company usage through OpenRouter suggests restrictions could affect existing production choices, not only future experimentation.
The trend: AI model consumption is shifting toward a multi-model, inference-driven market in which reasoning performance, open-weight availability, and routing flexibility jointly shape buyer choice.
Most interesting chart in this for me so far is this one: I'm surprised about the the large MiniMax M2 share—this is not a model I hear much about. [image]
We collaborated with @a16z to publish the **State of AI** - an empirical report on how LLMs have been used on OpenRouter. After analyzing more than 100 trillion tokens across hundreds of models and 3+ million users (excluding 3rd party) from the last year, we have a lot of [image…
We hope that it will help you make better decisions about where to invest, what to build, and where AI adoption is heading next. Read the paper here https://openrouter.ai/...
OSS isn't “just for tinkering” - it is extremely popular in two areas: 🧙♂️ Roleplay / creative dialogue: >50% of OSS usage 🧑💻 Programming assistance: ~15-20% [image]
The moment open-source models were close to 30% of OpenRouter traffic and almost all of them came from China with the notable models being: DeepSeek V3/R1, Qwen3 family, Kimi-K2 and GLM-4.5 + Air Minimax M2 is now also a major player, but open-weights models token-usage [image]
this one chart explains EVERYTHING about why OpenAI, xAI and Deepmind dropped everything to go chase after the grand prize in koding usecases as i said at AIE CODE and in my cogpost, Code AGI will be achieved in 20% of the time of full AGI, and capture 80% of the value of AGI. [i…
quite rich report from openrouter. points worth caring: - oss models have grown to have roughly 30% share on openrouter - code and companionship still major use-cases, ~70-80% - remaining use cases are like translation, trivia, general knowledge questions - chinese models have [i…
@AnjneyMidha ... One finding: we observe a Cinderella “Glass Slipper” effect for new models. Early users a new LLM either churn quickly or become part of a foundational cohort, with much higher retention than others. They are early adopters who can “lead” the rest of the market (…
>100 trillion token analysis of reasoning model usage over time Full piece from @MaikaThoughts, @AnjneyMidha, @xanderatallah, and @cclark: https://openrouter.ai/... [image]
On a prompt count basis this mean reasoning models are not close to a majority on OpenRouter, as reasoning models can use 10-1000x the tokens of non-thinking models per prompt. Lots of need for fast, efficient open models. Reasoning model usage is likely closed labs more. [image]
Lots of interesting details in this new report on usage trends from OpenRouter. openrouter.ai/state-of-ai I've been wondering about mean coding input token length, in their data it's around 20k tokens. Other large categories (roleplay, technology science) average around 5k [imag…