Daily snapshot · immutable permalink
AI model usage rankings - July 21, 2026
As published on July 21, 2026, Xiaomi: MiMo-V2.5 led real OpenRouter usage share at 17.6%. OpenAI: GPT-5.2 (batch), Poolside: Laguna S 2.1 (free), inclusionAI: Ling-2.6-flash, Qwen: Qwen3.7 Plus entered the leaderboard for the first time. OpenAI: GPT-5.4 (batch) moved 12 places up since July 20, 2026.
- #1 Xiaomi: MiMo-V2.5 - 17.6% usage share
- #2 DeepSeek: DeepSeek V4 Flash 0423 - 10.2% usage share
- #3 Tencent: Hy3 - 7.1% usage share
- #4 Z.ai: GLM 5.2 (batch) - 6.8% usage share
- #5 DeepSeek: DeepSeek V4 Pro - 4.9% usage share
- #6 Anthropic: Claude Opus 4.8 (batch) - 4.8% usage share
- #7 MiniMax: MiniMax M3 (batch) - 4.2% usage share
- #8 StepFun: Step 3.7 Flash - 3.7% usage share
- #9 NVIDIA: Nemotron 3 Ultra (free) - 2.6% usage share
- #10 MoonshotAI: Kimi K3 - 2.3% usage share
| # | Model | Usage share | Rank change |
|---|---|---|---|
| 01 | Xiaomi: MiMo-V2.5 | 17.6% | +1 (+1.9pts) |
| 02 | DeepSeek: DeepSeek V4 Flash 0423 | 10.2% | +1 (+1.8pts) |
| 03 | Tencent: Hy3 | 7.1% | -2 (-13.7pts) |
| 04 | Z.ai: GLM 5.2 (batch) | 6.8% | +1 (+2pts) |
| 05 | DeepSeek: DeepSeek V4 Pro | 4.9% | -1 (-0.2pts) |
| 06 | Anthropic: Claude Opus 4.8 (batch) | 4.8% | +3 (+2.3pts) |
| 07 | MiniMax: MiniMax M3 (batch) | 4.2% | -1 (+1.1pts) |
| 08 | StepFun: Step 3.7 Flash | 3.7% | +6 (+2.2pts) |
| 09 | NVIDIA: Nemotron 3 Ultra (free) | 2.6% | -2 (-0.4pts) |
| 10 | MoonshotAI: Kimi K3 | 2.3% | +1 (+0.4pts) |
| 11 | Anthropic: Claude Opus 4.7 (batch) | 2.2% | -3 (-0.3pts) |
| 12 | Anthropic: Claude Sonnet 5 (batch) | 2.1% | -2 (+0.1pts) |
| 13 | Google: Gemini 3 Flash Preview (batch) | 1.9% | -1 (+0.1pts) |
| 14 | Anthropic: Claude Sonnet 4.6 (batch) | 1.6% | -1 (+0.1pts) |
| 15 | Google: Gemini 3.1 Flash Lite (batch) | 1.1% | no change |
| 16 | Xiaomi: MiMo-V2.5-Pro | 1.1% | +4 (+0.1pts) |
| 17 | Google: Gemini 3.5 Flash (batch) | 1.1% | +6 (+0.3pts) |
| 18 | Google: Gemini 2.5 Flash Lite (batch) | 1.1% | -1 (+0.1pts) |
| 19 | Google: Gemini 2.5 Flash (batch) | 1.1% | no change |
| 20 | OpenAI: GPT-5.6 Sol (batch) | 1.0% | -2 (-0.1pts) |
| 21 | OpenAI: GPT-5.5 (batch) | 0.9% | -5 (-0.1pts) |
| 22 | SpaceXAI: Grok 4.5 | 0.9% | +7 (+0.3pts) |
| 23 | OpenAI: GPT-5.4 (batch) | 0.8% | +12 (+0.5pts) |
| 24 | OpenAI: gpt-oss-120b | 0.8% | -3 (0pts) |
| 25 | NVIDIA: Nemotron 3 Super (free) | 0.7% | +6 (+0.2pts) |
| 26 | Google: Gemma 4 31B (free) | 0.7% | -1 (+0.1pts) |
| 27 | DeepSeek: DeepSeek V3.2 | 0.6% | -3 (0pts) |
| 28 | OpenAI: GPT-5.6 Luna (batch) | 0.6% | no change |
| 29 | Anthropic: Claude Fable 5 (batch) | 0.6% | -7 (-0.2pts) |
| 30 | Google: Gemma 4 26B A4B (free) | 0.6% | -3 (0pts) |
| 31 | MoonshotAI: Kimi K2.6 | 0.6% | -5 (0pts) |
| 32 | OpenAI: GPT-4o-mini (batch) | 0.6% | -2 (0pts) |
| 33 | Mistral: Mistral Nemo | 0.6% | +1 (+0.2pts) |
| 34 | Anthropic: Claude Haiku 4.5 (batch) | 0.4% | -2 (0pts) |
| 35 | Anthropic: Claude Opus 4.6 (batch) | 0.4% | -2 (0pts) |
| 36 | OpenAI: GPT-5.2 (batch) New | 0.4% | - |
| 37 | OpenAI: GPT-5.4 Mini (batch) | 0.4% | +2 (+0.1pts) |
| 38 | Google: Gemini 3.1 Pro Preview (batch) | 0.4% | -1 (+0.1pts) |
| 39 | Poolside: Laguna S 2.1 (free) New | 0.3% | - |
| 40 | OpenAI: GPT-5.4 Nano (batch) | 0.3% | -4 (0pts) |
| 41 | OpenAI: GPT-5.6 Terra (batch) | 0.3% | -3 (0pts) |
| 42 | OpenAI: GPT-5 Mini (batch) | 0.3% | -2 (0pts) |
| 43 | inclusionAI: Ling-2.6-flash New | 0.2% | - |
| 44 | Qwen: Qwen3.7 Plus New | 0.2% | - |
| 45 | MoonshotAI: Kimi K2.7 Code (batch) | 0.2% | no change |
This page is a permanent, unchanging record of real OpenRouter usage share as it stood on July 21, 2026. Later corrections to the data pipeline appear as new dated snapshots - this page's figures never change.
Source: OpenRouter (openrouter.ai/rankings), as of July 21, 2026.
Benchmark scores: Artificial Analysis (artificialanalysis.ai) via OpenRouter (openrouter.ai/rankings).
Methodology: smophy.ai/benchmark/methodology
Token counts originate from each provider's own tokenizer and are not directly comparable across providers.