CodeSOTAIntelligence
Demand Research · ort.fabryka.ai

Who Burns the Tokens: OpenRouter Apps

Models are the supply side; apps are the demand. This maps the ~45 ranked apps to the models and vendors they run on — revealing how concentrated each vendor's volume is on a handful of agent apps.

snapshot 2026-09-25 · top-20 model breakdown per app · generated 2026-09-25 · ← dashboard · market memo

Token flow — apps → vendors → origin

Where the demand actually goes: each top app's tokens flowing to the vendors it runs on, then to the model's country of origin. Ribbon width = tokens; red = Chinese-origin, blue = Western.

Hermes AgentClaude CodeKilo CodeClinepiCodexOpenClawOpenHandsISEKAI ZEROPortkey AIdeepseekz-aiOther vendorsopenaiupstagepoolsideminimaxanthropicnvidiaChinaWestOther

Token flow — apps → the actual models

The same demand, resolved to the concrete models each app routes to (top 9 models; the rest grouped as "Other models").

Hermes AgentClaude CodeKilo CodeClinepiCodexOpenClawOpenHandsISEKAI ZEROPortkey AIOther modelsGLM 5.3 Flash (batch)DeepSeek V4 Flash 0731DeepSeek V4.1 Flash (batDeepSeek V4 Flash 0423Solar Pro 4Laguna S 2.1 (free)MiniMax M3GLM 5.3 (batch)GPT-5.6 Luna (batch)

The numbers behind the flow

Top apps by token volume (T)
Demand mix per app — 🔴 Chinese · 🔵 Western · ⚪ other (% of the app's tokens)
Vendor concentration — top-app share of each vendor's app volume (red = ≥40% concentration risk)

1 · App leaderboard — who consumes the most

apptokensshareruns on (vendor mix)
Hermes Agent40.7T35.7%deepseek 45% · z-ai 17% · upstage 10%
Claude Code19.8T17.4%z-ai 57% · anthropic 14% · deepseek 9%
Kilo Code12.2T10.6%poolside 22% · nvidia 20% · dots-studio 12%
Cline11.7T10.2%z-ai 35% · deepseek 31% · meta 11%
pi7.3T6.4%deepseek 44% · z-ai 27% · openai 10%
Codex4.9T4.3%openai 79% · deepseek 7% · meta 6%
OpenClaw3.9T3.4%deepseek 42% · z-ai 19% · minimax 11%
OpenHands1.6T1.4%deepseek 71% · google 11% · z-ai 5%
ISEKAI ZERO1.6T1.4%xiaomi 48% · z-ai 22% · deepseek 22%
Portkey AI1.2T1.1%google 48% · z-ai 18% · qwen 8%
Lemonade1.1T1.0%openai 66% · google 15% · z-ai 9%
Descript1.1T1.0%anthropic 77% · google 20% · z-ai 2%
Nous Research API1.0T0.9%deepseek 33% · poolside 20% · z-ai 14%
Janitor AI782.8B0.7%deepseek 60% · z-ai 15% · nvidia 13%
Open WebUI677.5B0.6%openai 64% · deepseek 16% · z-ai 9%
Zed Editor609.5B0.5%deepseek 38% · z-ai 35% · openai 7%
LangChain484.1B0.4%deepseek 33% · openai 25% · google 17%
Craft481.3B0.4%meta 36% · deepseek 17% · google 15%
Mira451.8B0.4%minimax 62% · openai 29% · z-ai 4%
GDevelop395.8B0.3%deepseek 37% · z-ai 22% · meta 22%
Ito387.9B0.3%openai 100% · google 0% · anthropic 0%
SillyTavern277.7B0.2%deepseek 34% · z-ai 28% · google 19%
Stanley262.4B0.2%google 49% · minimax 35% · anthropic 16%
extra.email173.7B0.2%deepseek 60% · z-ai 36% · typesafe 3%
Roo Code173.5B0.2%deepseek 46% · z-ai 15% · minimax 12%

Share = % of tokens across all tracked apps. "Runs on" = the app's top vendors by token volume.

2 · Vendor demand concentration — who depends on whom

vendortokens (via apps)top-app conc.top driving apps
deepseek33.4T55%Hermes Agent 55% · Cline 11% · pi 10%
z-ai27.4T41%Claude Code 41% · Hermes Agent 25% · Cline 15%
openai8.9T44%Codex 44% · Hermes Agent 18% · Lemonade 8%
anthropic5.4T51%Claude Code 51% · Descript 15% · Hermes Agent 14%
upstage5.0T86%Hermes Agent 86% · Cline 12% · Nous Research API 1%
minimax4.9T36%Hermes Agent 36% · Kilo Code 25% · Claude Code 16%
poolside4.7T57%Kilo Code 57% · Hermes Agent 27% · Cline 12%
nvidia4.0T62%Kilo Code 62% · Hermes Agent 26% · OpenClaw 7%
google3.7T19%Claude Code 19% · Portkey AI 16% · Hermes Agent 15%
tencent2.9T41%Hermes Agent 41% · Claude Code 37% · Kilo Code 16%

Top-app conc. = share of the vendor's app-driven volume from its single biggest app — high (≥40%) = concentration risk, low (<20%) = broad, durable demand.

3 · Why it matters