FrontierCode 2.0: Claude Opus 5.5 tops private coding benchmark at 46.4%, Kimi and DeepSeek close behind
bookwormengr · x · 2026-10-28
Artificial Analysis launched Cognition's FrontierCode 2.0, a private agentic coding benchmark of 132 real open-source repo tasks graded on whether a maintainer would merge the change. Top scores: Claude Opus 5.5 (Max) 46.4%, Opus 5.5 (Xhigh) 43.8%, GPT-6 Astra (Max) 42.4%.
Author bookwormengr highlights that DeepSeek 4.1 Flash 1017 (Max) sits just behind Grok 4.7 (xHigh) and Claude 5.5 Fable (Max), while Kimi K3.1 Pro trails Google Aragon (high). He rebuts Lisan's earlier claim that Kimi 3.1 and DS v4.1 were absent from the board, noting Artificial Analysis confirmed Kimi 3.1 outperforms Muse Wonder (43) and nearly matches Fable (49) — results just withheld on private benchmarks. The full chart runs down to Thinking Machines' Inklings 1.0 at 4%.
Related event: FrontierCode 2.0 Benchmark: OpenAI and Anthropic Dominate Top 20(2 posts)→
More from Models
- Cryptic insider tease sparks guesses: GPT-6.2 Sol or GPT-6.1 Astra? — CtrlAltDwayne · 2026-10-28
- ECI Metric Adopted by Anthropic, Google and Artificial Analysis — Jsevillamol · 2026-10-28
- 'The OpenAI Canyon' meme chart shows OpenAI models trailing on output tokens — steipete · 2026-10-28
- First open-weight production dLLM: Clover-35B hits 680 tps on one GPU, beats same-size AR models on SWE-bench — taoyds · 2026-10-28
- Only Opus 5.5 one-shots "Anthropic" in association test, revealing a persona shift — ryunuck · 2026-10-28
- ChatGPT Pro usage limits cut in half in 3 days, devs voice frustration at OpenAI — johnlindquist · 2026-10-28