Daily snapshot · immutable permalink
AI model usage rankings - August 19, 2026
As published on August 19, 2026, DeepSeek: DeepSeek V4 Flash 0423 led real OpenRouter usage share at 20.4%. Z.ai: GLM 5.3, Google: Gemini 3.5 Flash Lite (batch), OpenAI: GPT-5 Mini (batch) entered the leaderboard for the first time. Anthropic: Claude Opus 4.7 (batch) moved 20 places down since August 18, 2026.
- #1 DeepSeek: DeepSeek V4 Flash 0423 - 20.4% usage share
- #2 Tencent: Hy3 - 10.0% usage share
- #3 Xiaomi: MiMo-V2.5 - 9.7% usage share
- #4 OpenAI: GPT-5.6 Luna (batch) - 6.7% usage share
- #5 NVIDIA: Nemotron 3 Ultra (free) - 6.0% usage share
- #6 Z.ai: GLM 5.2 (free) - 4.6% usage share
- #7 DeepSeek: DeepSeek V4 Pro 0423 - 3.4% usage share
- #8 MiniMax: MiniMax M3 (free) - 2.6% usage share
- #9 Google: Gemini 3.7 Flash (batch) - 2.3% usage share
- #10 Poolside: Laguna S 2.1 (free) - 2.2% usage share
| # | Model | Usage share | Rank change |
|---|---|---|---|
| 01 | DeepSeek: DeepSeek V4 Flash 0423 | 20.4% | no change |
| 02 | Tencent: Hy3 | 10.0% | no change |
| 03 | Xiaomi: MiMo-V2.5 | 9.7% | no change |
| 04 | OpenAI: GPT-5.6 Luna (batch) | 6.7% | no change |
| 05 | NVIDIA: Nemotron 3 Ultra (free) | 6.0% | +1 (+2pts) |
| 06 | Z.ai: GLM 5.2 (free) | 4.6% | -1 (+0.2pts) |
| 07 | DeepSeek: DeepSeek V4 Pro 0423 | 3.4% | no change |
| 08 | MiniMax: MiniMax M3 (free) | 2.6% | no change |
| 09 | Google: Gemini 3.7 Flash (batch) | 2.3% | +1 (+0.5pts) |
| 10 | Poolside: Laguna S 2.1 (free) | 2.2% | -1 (+0.3pts) |
| 11 | Claude Opus 5 (batch) | 2.1% | no change |
| 12 | OpenAI: GPT-5.6 Sol (batch) | 1.9% | no change |
| 13 | MoonshotAI: Kimi K3 | 1.7% | no change |
| 14 | Anthropic: Claude Sonnet 5 (batch) | 1.5% | no change |
| 15 | Google: Gemini 3 Flash Preview (batch) | 1.2% | +1 (+0.1pts) |
| 16 | NVIDIA: Nemotron 3.5 Lightning (free) | 1.1% | +2 (0pts) |
| 17 | Anthropic: Claude Sonnet 4.6 (batch) | 1.0% | +6 (+0.2pts) |
| 18 | OpenAI: GPT-5.6 Terra (batch) | 1.0% | -1 (-0.1pts) |
| 19 | Google: Gemini 2.5 Flash Lite (batch) | 0.9% | +1 (+0.1pts) |
| 20 | Xiaomi: MiMo-V2.5-Pro | 0.8% | +4 (0pts) |
| 21 | Upstage: Solar Pro 4 | 0.7% | +6 (0pts) |
| 22 | Google: Gemini 3.1 Pro Preview (batch) | 0.6% | +14 (+0.3pts) |
| 23 | Google: Gemini 2.5 Flash (batch) | 0.6% | +8 (+0.1pts) |
| 24 | Google: Gemini 3.1 Flash Lite (batch) | 0.6% | +5 (0pts) |
| 25 | StepFun: Step 3.7 Flash | 0.6% | -3 (-0.2pts) |
| 26 | SpaceXAI: Grok 4.6 | 0.6% | no change |
| 27 | OpenAI: GPT-5.6 Luna Pro (batch) | 0.6% | -8 (-0.4pts) |
| 28 | NVIDIA: Nemotron 3 Super (free) | 0.6% | +2 (0pts) |
| 29 | OpenAI: gpt-oss-120b | 0.6% | -4 (-0.1pts) |
| 30 | Anthropic: Claude Opus 4.8 (batch) | 0.6% | +2 (0pts) |
| 31 | Z.ai: GLM 5.3 New | 0.5% | - |
| 32 | Google: Gemini 3.6 Flash (batch) | 0.5% | -17 (-0.7pts) |
| 33 | Google: Gemma 4 31B (free) | 0.5% | +2 (+0.1pts) |
| 34 | DeepSeek: DeepSeek V3.2 | 0.5% | -1 (0pts) |
| 35 | Google: Gemma 4 26B A4B (free) | 0.4% | -1 (0pts) |
| 36 | OpenAI: GPT-5.5 (batch) | 0.4% | -8 (-0.2pts) |
| 37 | Qwen: Qwen3.8 Max | 0.4% | +1 (+0.1pts) |
| 38 | Anthropic: Claude Fable 5 (batch) | 0.4% | -1 (0pts) |
| 39 | Anthropic: Claude Haiku 4.5 (batch) | 0.3% | no change |
| 40 | OpenAI: GPT-4o-mini (batch) | 0.3% | no change |
| 41 | Anthropic: Claude Opus 4.7 (batch) | 0.3% | -20 (-0.6pts) |
| 42 | Google: Gemini 3.5 Flash Lite (batch) New | 0.3% | - |
| 43 | MiniMax: MiniMax M2.7 (free) | 0.3% | +1 (0pts) |
| 44 | Qwen: Qwen3.7 Flash | 0.3% | +1 (0pts) |
| 45 | OpenAI: GPT-5.4 (batch) | 0.2% | -2 (0pts) |
| 46 | Mistral: Mistral Nemo | 0.2% | no change |
| 47 | OpenAI: GPT-5 Mini (batch) New | 0.2% | - |
This page is a permanent, unchanging record of real OpenRouter usage share as it stood on August 19, 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 August 19, 2026.
Benchmark scores: Artificial Analysis (artificialanalysis.ai) via OpenRouter (openrouter.ai/rankings).
Methodology: www.smophy.ai/benchmark/methodology
Token counts originate from each provider's own tokenizer and are not directly comparable across providers.