Muse GlimmervsGPT-5.6 Luna
Muse Glimmer | GPT-5.6 Luna | |
|---|---|---|
| Specifications | ||
Parameters | 30B | — |
| Benchmarks | ||
Nonsense detection BullshitBench v2Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better. | — | 40% |
Prompt injection robustness Gray Swan IPI · k = 1Attackers hide malicious instructions inside content the AI reads — a web page, an email, a document — and try to hijack what it does. Gray Swan's indirect prompt injection benchmark measures how often such an attack succeeds when the attacker gets a single try. Lower is better. | — | 8.3% |
Prompt injection robustness Gray Swan IPI · k = 10Attackers hide malicious instructions inside content the AI reads — a web page, an email, a document — and try to hijack what it does. This variant gives the attacker 10 tries and counts an attack as successful if any of them works. Lower is better. | — | 38.6% |
Prompt injection robustness Gray Swan IPI · k = 15Attackers hide malicious instructions inside content the AI reads — a web page, an email, a document — and try to hijack what it does. This variant gives the attacker 15 tries and counts an attack as successful if any of them works. Lower is better. | — | 43.9% |
Agentic coding SWE-Bench ProCan the AI fix real bugs in real software? It's handed actual problems from open-source projects and has to write code that genuinely solves them. Higher is better. | 51.2% | — |
Coding SWE-Bench VerifiedReal coding tasks pulled from open-source projects — the AI has to find and fix actual bugs. A human-checked version of the original SWE-Bench. Higher is better. | 76% | — |
Agentic coding CursorBench v3.2Cursor's own test of harder, real-world coding tasks inside a code editor, on the refreshed v3.2 task set. Scores aren't comparable with v3.1. Higher is better. | — | 61.1% |
Agentic computer work Frontier-Bench v0.1A hard, ever-evolving set of real computer tasks — coding, system administration, data work, and more — that an AI agent has to complete on its own. Run by the Harbor / Laude Institute team as the successor to Terminal-Bench (v0.1 is the first release of the task set). The score is the share of tasks solved. Higher is better. | — | 14.3% |
Agentic terminal coding Terminal-Bench 2.1Can the AI work in a command-line terminal — running commands and finishing technical setup tasks the way a developer would? Higher is better. | 51.7% | 82.5%Best |
Multi-step tool use MCP AtlasCan the AI chain together many tools and steps to complete one bigger task, rather than doing just a single thing? Higher is better. | 75.5% | — |
Multidisciplinary reasoning Humanity's Last Exam · no toolsHumanity's Last Exam — extremely hard expert questions across many subjects, written so you can't just look up the answer. “No tools” means the AI answers on its own. Higher is better. | 22% | — |
Science GPQA DiamondGraduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | 83.5% | — |
Agentic computer use OSWorld-VerifiedCan the AI actually operate a computer — clicking, typing, and using real apps — to finish tasks on its own? Higher is better. | 65.9% | — |
Knowledge work GDPval-AA v2economically valuable knowledge work (v2, re-based Elo) | 953 | — |
Chart reasoning CharXiv ReasoningCan the AI read and reason about complex charts and figures, not just text? Higher is better. | 78.8% | — |
Multimodal reasoning MMMU-ProA tougher version of MMMU — college-level questions that mix images, diagrams, and text together. Higher is better. | 74% | — |
Community preference (code) Arena Elo (Code)Like the text arena, but people vote on which AI writes better code. The votes become a chess-style Elo rating on arena.ai. Higher is better. | — | 1523 |
| Overview | ||
| Company | Meta | OpenAI |
| Release date | Aug 10 2026 | Jun 26 2026 |
| Access | Open Weight | Proprietary |
Which is better: Muse Glimmer or GPT-5.6 Luna?
GPT-5.6 Luna leads Muse Glimmer on 1 of the 1 benchmark they both report (Terminal-Bench 2.1). GPT-5.6 Luna shipped 45 days before Muse Glimmer, so benchmark comparisons should account for the intervening progress.
Muse Glimmer is open weight, while GPT-5.6 Luna is proprietary.
On Terminal-Bench 2.1, GPT-5.6 Luna leads at 82.5% vs Muse Glimmer at 51.7%.
Frequently asked questions
Muse Glimmer was released by Meta on Aug 10 2026.
GPT-5.6 Luna was released by OpenAI on Jun 26 2026.
GPT-5.6 Luna leads on Terminal-Bench 2.1 — Muse Glimmer 51.7% vs GPT-5.6 Luna 82.5%.
Muse Glimmer is an open weight model released by Meta. GPT-5.6 Luna is a proprietary model released by OpenAI.
Other comparisons
Muse GlimmervsClaude Opus 5GPT-5.6 LunavsClaude Opus 5Muse GlimmervsGemini 3.6 FlashGPT-5.6 LunavsGemini 3.6 FlashMuse GlimmervsGrok 4.5GPT-5.6 LunavsGrok 4.5Muse GlimmervsDeepSeek-V4-Flash-0731GPT-5.6 LunavsDeepSeek-V4-Flash-0731Muse GlimmervsMistral Medium 3.5GPT-5.6 LunavsMistral Medium 3.5Muse GlimmervsKimi K3GPT-5.6 LunavsKimi K3