webAI Air-26B-A4B Card - MMLU-Pro (2,000-question sample): leaderboard
MMLU-Pro ten-option knowledge and reasoning questions, a 2,000-question sample spread evenly across subjects (of 12,032). Thinking on, temperature 1.0, 16,384-token answer cap. Score (%). From webAI's Air-26B-A4B model card (November 4, 2026): Google's BF16 Gemma 4 26B A4B (IT) against webAI's 4-bit Air quantization of it, both run by webAI on the same items with the same settings, one sampled run each (Air with an 8-bit KV cache, BF16 with a 16-bit one). Higher is better.
Metric: Score (%). Source: huggingface.co. Saturation forecast: Around February 2027. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 4 26B A4B (IT) (Reasoning) | 82.8 | #545 (Gemma 4 26B A4B (IT)) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=webai-air-26b-a4b-card-mmlu-pro-2-000-question-sample · How It Works · Data refreshed daily, snapshot 2026-11-12.