Daily snapshot · immutable permalink
AI model usage rankings - September 30, 2026
As published on September 30, 2026, Space Bunny Alpha led real OpenRouter usage share at 21.9%. OpenAI: GPT-6.1 Sol (batch), SpaceXAI: Grok 4.7, DeepSeek: DeepSeek V3.2, NVIDIA: Nemotron 3 Super (free) entered the leaderboard for the first time. OpenAI: GPT-6 Astra (batch) moved 9 places up since September 29, 2026.
- #1 Space Bunny Alpha - 21.9% usage share
- #2 DeepSeek: DeepSeek V4.1 Flash (batch) - 14.3% usage share
- #3 Xiaomi: MiMo-V2.6-Flash - 6.0% usage share
- #4 DeepSeek: DeepSeek V4 Flash 0423 - 5.9% usage share
- #5 Z.ai: GLM 5.3 Flash (batch) - 5.8% usage share
- #6 OpenAI: GPT-6 Luna (batch) - 4.4% usage share
- #7 Tencent: Hy4 preview - 4.3% usage share
- #8 NVIDIA: Nemotron 3 Ultra (free) - 2.8% usage share
- #9 OpenAI: GPT-5.6 Luna (batch) - 2.8% usage share
- #10 Anthropic: Claude Opus 5.5 (batch) - 1.8% usage share
| # | Model | Usage share | Rank change |
|---|---|---|---|
| 01 | Space Bunny Alpha | 21.9% | no change |
| 02 | DeepSeek: DeepSeek V4.1 Flash (batch) | 14.3% | no change |
| 03 | Xiaomi: MiMo-V2.6-Flash | 6.0% | +1 (+0.3pts) |
| 04 | DeepSeek: DeepSeek V4 Flash 0423 | 5.9% | +1 (+0.2pts) |
| 05 | Z.ai: GLM 5.3 Flash (batch) | 5.8% | -2 (-0.2pts) |
| 06 | OpenAI: GPT-6 Luna (batch) | 4.4% | +2 (+1.2pts) |
| 07 | Tencent: Hy4 preview | 4.3% | no change |
| 08 | NVIDIA: Nemotron 3 Ultra (free) | 2.8% | +1 (0pts) |
| 09 | OpenAI: GPT-5.6 Luna (batch) | 2.8% | -3 (-2.3pts) |
| 10 | Anthropic: Claude Opus 5.5 (batch) | 1.8% | +3 (+0.6pts) |
| 11 | Tencent: Hy3 | 1.6% | -1 (-0.1pts) |
| 12 | Z.ai: GLM 5.3 (batch) | 1.6% | -1 (0pts) |
| 13 | Google: Gemini 3.8 Flash (batch) | 1.5% | -1 (+0.2pts) |
| 14 | OpenAI: GPT-5.6 Sol (batch) | 1.1% | +8 (+0.4pts) |
| 15 | MoonshotAI: Kimi K3 (batch) | 1.1% | -1 (0pts) |
| 16 | DeepSeek: DeepSeek V4 Pro 0423 | 0.9% | -1 (0pts) |
| 17 | Meta: Muse Spark 1.3 Contributor | 0.9% | no change |
| 18 | Xiaomi: MiMo-V2.6-Pro | 0.9% | +3 (+0.1pts) |
| 19 | Z.ai: GLM 5.2 | 0.8% | +1 (+0.1pts) |
| 20 | MiniMax: MiniMax M3 | 0.8% | -1 (0pts) |
| 21 | Anthropic: Claude Sonnet 5 (batch) | 0.8% | -3 (0pts) |
| 22 | Upstage: Solar Pro 4 | 0.7% | +1 (0pts) |
| 23 | OpenAI: GPT-6 Sol (batch) | 0.7% | -7 (-0.2pts) |
| 24 | Xiaomi: MiMo-V2.5 | 0.7% | no change |
| 25 | OpenAI: GPT-6 Astra (batch) | 0.6% | +9 (+0.3pts) |
| 26 | Anthropic: Claude Sonnet 5.5 (batch) | 0.6% | +3 (+0.2pts) |
| 27 | Poolside: Laguna S 2.1 (free) | 0.6% | -2 (0pts) |
| 28 | Google: Gemini 2.5 Flash Lite (batch) | 0.4% | -2 (0pts) |
| 29 | OpenAI: GPT-6.1 Sol (batch) New | 0.4% | - |
| 30 | OpenAI: gpt-oss-120b (batch) | 0.4% | -2 (0pts) |
| 31 | Google: Gemini 3 Flash Preview (batch) | 0.4% | -4 (0pts) |
| 32 | Google: Gemini 3.1 Flash Lite (batch) | 0.3% | no change |
| 33 | OpenAI: GPT-6 Luna Pro (batch) | 0.3% | +9 (+0.1pts) |
| 34 | Qwen: Qwen3.8 Flash | 0.3% | -4 (-0.1pts) |
| 35 | Google: Gemini 3.7 Flash (batch) | 0.3% | -2 (0pts) |
| 36 | Google: Gemini 2.5 Flash (batch) | 0.3% | -1 (0pts) |
| 37 | Google: Gemma 4 31B (free) | 0.3% | +2 (0pts) |
| 38 | NVIDIA: Nemotron 3.5 Lightning (free) | 0.3% | -2 (0pts) |
| 39 | Anthropic: Claude Opus 5 (batch) | 0.3% | -8 (-0.1pts) |
| 40 | Qwen: Qwen3.8 27B | 0.3% | +5 (+0.1pts) |
| 41 | Anthropic: Claude Fable 5.1 (batch) | 0.3% | -4 (0pts) |
| 42 | inclusionAI: Ling 3.0 Flash Sante | 0.2% | -2 (0pts) |
| 43 | SpaceXAI: Grok 4.7 New | 0.2% | - |
| 44 | DeepSeek: DeepSeek V3.2 New | 0.2% | - |
| 45 | NVIDIA: Nemotron 3 Super (free) New | 0.2% | - |
| 46 | Anthropic: Claude Sonnet 4.6 (batch) | 0.2% | -2 (0pts) |
This page is a permanent, unchanging record of real OpenRouter usage share as it stood on September 30, 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 September 30, 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.
