DeepSeek's DeepSeek V4 Flash leads OpenRouter with 5.35T tokens (6 Jul 2026)

Steady week on OpenRouter. DeepSeek's DeepSeek V4 Flash holds the top of the leaderboard, ahead of Tencent's Hy3 preview and DeepSeek's DeepSeek V4 Pro. The top three on each ranking stayed intact this week. Smaller reshuffles below: see the per-area breakdowns for the movers.

LLM Leaderboard

The big picture: which models AI builders are paying to run this week.

1. DeepSeek V4 Flash by DeepSeek: 5.35T tokens ↑14%
2. Hy3 preview by Tencent: 3.37T tokens ↑3%
3. DeepSeek V4 Pro by DeepSeek: 2.25T tokens ↑10%
4. Claude Opus 4.8 by Anthropic: 2.08T tokens ↑6%
5. Claude Opus 4.7 by Anthropic: 2.02T tokens ↑14%

Market Share

Vendor share of total OpenRouter traffic: who's winning the platform race.

1. google 876M tokens ↑28.1%
2. deepseek 640M tokens ↑20.5%
3. openai 507M tokens ↑16.3%
4. qwen 182M tokens ↑5.8%
5. anthropic 164M tokens ↑5.3%

Benchmark Leaders (AA Index)

Independent benchmark scores via Artificial Analysis. High AA Index = stronger reasoning, but cost matters too.

1. Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) by Anthropic: 59.9 AA Index
2. Claude Opus 4.8 (Adaptive Reasoning, Max Effort) by Anthropic: 55.7 AA Index
3. GPT-5.5 (xhigh) by OpenAI: 54.8 AA Index
4. Claude Opus 4.7 (Adaptive Reasoning, Max Effort) by Anthropic: 53.5 AA Index
5. Claude Sonnet 5 (Adaptive Reasoning, Max Effort) by Anthropic: 53.4 AA Index

Fastest Models

Throughput champs: pick these for latency-sensitive apps where speed beats raw quality.

1. gpt-oss-120b
2. gpt-oss-20b
3. gpt-oss-safeguard-20b
4. Qwen3 32B
5. Mercury 2

Top Models for English

Most-used models for English content this week: the multilingual leaders.

1. DeepSeek V4 Flash by DeepSeek: 863B tokens ↑13.7%
2. Hy3 preview by Tencent: 332B tokens ↑5.3%
3. DeepSeek V4 Pro by DeepSeek: 324B tokens ↑5.2%
4. GPT-5.5 by OpenAI: 163B tokens ↑2.6%
5. Claude Sonnet 4.6 by Anthropic: 155B tokens ↑2.5%

Top Models for Python

Which models are getting picked for Python work right now.

1. DeepSeek V4 Flash by DeepSeek: 216B tokens ↑13.8%
2. DeepSeek V4 Pro by DeepSeek: 88B tokens ↑5.6%
3. Hy3 preview by Tencent: 88B tokens ↑5.6%
4. Claude Opus 4.7 by Anthropic: 37.4B tokens ↑2.4%

Top Models for short prompts (1K-10K tokens)

For short prompts (1K-10K tokens, the bulk of typical traffic), here's what builders chose.

1. DeepSeek V4 Flash by DeepSeek: 271M requests ↑17.5%
2. Gemini 2.5 Flash Lite by Google: 111M requests ↑7.1%
3. Gemini 2.5 Flash by Google: 94.7M requests ↑6.1%
4. gpt-oss-120b by OpenAI: 71.6M requests ↑4.6%
5. Gemini 3 Flash Preview by Google: 68.6M requests ↑4.4%

Top Models for Tool Calls

If your stack uses tool calls / function calling, these models are getting the most invocations.

1. DeepSeek V4 Flash by DeepSeek: 67.5M tokens ↑11.9%
2. Hy3 preview by Tencent: 34.5M tokens ↑6.1%
3. DeepSeek V4 Pro by DeepSeek: 25M tokens ↑4.4%
4. GPT-4.1 Mini by OpenAI: 21.7M tokens ↑3.8%
5. Gemini 3 Flash Preview by Google: 19.8M tokens ↑3.5%

Top Image Models

Image-generation through OpenRouter: most-served models this week.

1. Gemini 2.5 Flash Lite by Google: 186M images ↑26.4%
2. Gemini 2.5 Flash by Google: 70M images ↑9.9%
3. Qwen3 VL 235B A22B Instruct by Qwen: 42.3M images ↑6%
4. Gemini 3 Flash Preview by Google: 40.5M images ↑5.7%
5. Claude Sonnet 4.6 by Anthropic: 29.2M images ↑4.1%

Top Apps on OpenRouter

Useful as social proof when picking your stack: the largest public apps and agents that opt into OpenRouter usage tracking.

Most Popular

1. Hermes Agent (26.1T tokens): Hermes Agent is an open-source, self-improving AI agent by Nous Research that runs persistently with memory across sessions, and builds reusable skills from experience. It comes with 40+ built-in tools, including web search, browser automation, and vision, plus scheduled automations and subagents.
2. Kilo Code (7.38T tokens): Kilo Code is an open-source AI coding agent that works across VS Code, JetBrains, and CLI to help developers ship code faster with agentic workflows.
3. OpenClaw (4.96T tokens): OpenClaw is an open-source AI agent that connects to your messaging apps and takes real actions on your behalf, from running commands and browsing the web to managing files and sending emails.
4. Claude Code (4.16T tokens): Claude Code is Anthropic's agentic coding tool that reads your entire codebase, plans and executes changes across files, runs tests, and iterates on failures, all from natural language prompts.

Trending

Fastest-growing apps on OpenRouter this week: early signals of which builder workflows are breaking out.

1. Claude Code (1.22T tokens) ↑24%
2. Cline (514B tokens) ↑55%
3. OpenCX (95.8B tokens) ↑318%
4. Peezy Gateway (54.7B tokens) ↑100%
5. pi (542B tokens) ↑11%
6. Poolside
7. timely-router (51.3B tokens) ↑318%
8. LangChain (94.8B tokens) ↑48%

What this means for builders

The OpenRouter rankings are the cleanest signal of what AI builders pay to run, not what vendor marketing claims. This week the top of the leaderboard belongs to Google, and the fastest-growing app on the platform (Claude Code) is a useful tell about which builder workflows are quietly breaking out before the broader feed picks them up.