I have enough to answer with solid sourcing. Here's what I found. ## Answer **OpenCode Go doesn't officially designate a single "most advanced" model** — it's a rotating, curated lineup of open coding models, and the [official Go docs](https://opencode.ai/docs/go/) simply list them without ranking. That said, based on the lineup and independent benchmarks, the flagship/most-advanced tier is: **Kimi K3, Grok 4.7, GPT 6 Luna, GLM‑5.3, and Qwen3.8 Max**. If you want the single most advanced pick, it comes down to **Kimi K3 vs Grok 4.7**: - **Kimi K3** (Moonshot's flagship) is the standout "most advanced" *open* model — described as a 2.8‑trillion‑parameter, 3T‑class MoE with a **1M‑token context**, native video input, downloadable weights, and top-tier reasoning, which is why Go allocates it the stingiest usage (~110 requests/5h) [codingplan.org](https://codingplan.org/en/plans/opencode-go), [orcarouter.ai](https://www.orcarouter.ai/blog/grok-4-7-vs-kimi-k3). - **Grok 4.7** (xAI/SpaceXAI's newest flagship, released Sep 21 2026) actually edges it on independent numbers: on the **Artificial Analysis Intelligence Index v4.3.2 it scores 46 vs Kimi K3's 44**, and it's the **highest-ranked named OpenCode Go model on the AI Coding Daily leaderboard (#14, 52.28/60)** [orcarouter.ai](https://www.orcarouter.ai/blog/grok-4-7-vs-kimi-k3), [aicodingdaily.com/leaderboard](https://aicodingdaily.com/leaderboard). **On the full OpenCode Go lineup, here's how the top models rank on the AI Coding Daily benchmark (Sept 30, 2026):** | Model | Score /60 | Rank | |---|---|---| | Grok 4.7 (High) | 52.28 | #14 | | GPT‑6 Luna (Max) | 51.99 | #15 | | Kimi K3 (High) | 50.85 | #16 | | Grok 4.6 (High) | 50.15 | #17 | | GLM‑5.3 (High) | 49.92 | #20 | | Qwen 3.8 Max (High) | 48.7 | #22 | ([aicodingdaily.com](https://aicodingdaily.com/leaderboard)) **Caveats worth knowing:** - BenchLM's head-to-head rates **Kimi K3 higher than GPT‑6 Luna** overall (72.12 vs 64.95) and on coding (61.4 vs 50.2), but flags GPT‑6 Luna's evidence as "Estimated" [benchlm.ai](https://benchlm.ai/compare/gpt-6-luna-vs-kimi-k3). - The lineup **rotates frequently** — the official docs page (checked Oct 2026) lists 29 models, and models like Qwen3.7 Max, GLM‑5, and Kimi K2.5 have been dropped [opencode.ai/docs/go/](https://opencode.ai/docs/go/), [julien.cloud](https://julien.cloud/opencode-go-models/). **Bottom line:** For raw "most advanced," most subscribers treat **Kimi K3** as the flagship brain (biggest open model, 1M context), with **Grok 4.7** as the slightly higher-scoring, faster/cheaper frontier alternative. Note both are the most expensive models on Go, so they burn through your $12/5h allowance fastest.