Gemini 4 Argon: CI-First Benefit Score 6.3/10, which is CI-First Strong, with a Humics-Neutral protection badge. Google's first non-Flash model in seven months lands level with the top OpenAI model on the independent composite, first on the knowledge-work index and the best-calibrated model measured at its level, on a rate card that doubles on 2026-12-01 and access with no free tier. The frontier capability is real, and the discipline the score demands does not move with it.
The gap between open and closed AI models has both collapsed and widened. On composite benchmarks, the distance is 6 points. On crowdsourced preference, it is 29 Elo points. The answer to which side is winning depends entirely on which question you ask, and that is the most important finding of the year.
Grok 4.5: CI-First Benefit Score 5.8/10, Humics-Neutral Badge. Grok 4.5 is xAI's reasoning-capable large language model with a 500K token context window, multi-effort reasoning, and multimodal text-plus-image input. It serves as the balanced workhorse between Grok 4.3 and the newer Grok 4.6.
Qwen3.8 27B: CI-First Benefit Score 6.0/10, Humics-Neutral. Alibaba's 27B Mixture-of-Experts model balances reasoning, multilingual coverage, and local deployment for learners who need a capable model without frontier pricing.
GPT-5.6 Sol: CI-First Benefit Score 6.0/10, Humics-Neutral. OpenAI's flagship reasoning model with a 1M token context window, multi-modal inputs, and configurable reasoning effort. Released July 2026 at $4/M input and $20/M output tokens.
