Atlas

Benchmarks

← All benchmarks

ComplexConstraints (November 2026 rescoring)

Chat & Writing

Surge's ComplexConstraints board as rescored between its November 1 and November 3, 2026 site builds: instruction following in a single completion on the 75 prompts, with an LLM judge grading each rubric criterion. Neither the page nor the launch post states a protocol change, metric, judge or run count for these scores, so they are kept apart from the original board: GPT 6 Astra Max moved from 57.7 to 70.8 and GLM 5.3 Max from 30.5 to 54.4.

Benchmark source: Surgehq

Top models (higher is better)

ModelScoreSource
GPT-6 Astra70.8Surge AI ComplexConstraints (November 2026 rescoring) leaderboard · Surge AI
Gemini 4 Argon70.3Surge AI ComplexConstraints (November 2026 rescoring) leaderboard · Surge AI
GPT-6.1 Sol70.1Surge AI ComplexConstraints (November 2026 rescoring) leaderboard · Surge AI
Muse Wonder67.9Surge AI ComplexConstraints (November 2026 rescoring) leaderboard · Surge AI
Claude Sonnet 5.567.2Surge AI ComplexConstraints (November 2026 rescoring) leaderboard · Surge AI
Muse Spark 1.364.8Surge AI ComplexConstraints (November 2026 rescoring) leaderboard · Surge AI
Claude Opus 5.564.4Surge AI ComplexConstraints (November 2026 rescoring) leaderboard · Surge AI
Gemini 3.8 Flash63.2Surge AI ComplexConstraints (November 2026 rescoring) leaderboard · Surge AI
Grok 4.760.8Surge AI ComplexConstraints (November 2026 rescoring) leaderboard · Surge AI
Claude Fable 5.160.4Surge AI ComplexConstraints (November 2026 rescoring) leaderboard · Surge AI
GLM-5.354.4Surge AI ComplexConstraints (November 2026 rescoring) leaderboard · Surge AI
Kimi K351.1Surge AI ComplexConstraints (November 2026 rescoring) leaderboard · Surge AI
Mistral Large 4 Preview42.4Surge AI ComplexConstraints (November 2026 rescoring) leaderboard · Surge AI
Loading Atlas data…