Agents' Last Exam, Linux subset (Agents' Last Exam (Linux))
We show this table for reference; we do not rank on it.
The Linux subset of Agents' Last Exam, an end-to-end agent task benchmark, as reported by Meta.
Benchmark score on Agents' Last Exam (Linux) — November 11, 2026
We compile the Agents' Last Exam (Linux) rows from provider self-reports. Muse Ember leads the table at 61.6%, followed by GPT-6 Luna (57.1%) and Claude Haiku 5.5 (57.0%). We do not use these results to rank models overall.
Muse Ember
Meta
GPT-6 Luna
OpenAI
Claude Haiku 5.5
Anthropic
3 modelsAgenticCurrentDisplay onlyUpdated November 11, 2026
Benchmark score table (3 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Muse EmberMeta | 61.6% | Not reported | Closed |
| 2 | GPT-6 LunaOpenAI | 57.1% | Not reported | Closed |
| 3 | Claude Haiku 5.5Anthropic | 57.0% | Not reported | Closed |
Among the reported Agents' Last Exam (Linux) rows, Muse Ember is first at 61.6%. The third row is 4.6 points behind. The broader top-10 range is 4.6 points, so many of the published results sit in a relatively narrow band.
3 models have been evaluated on Agents' Last Exam (Linux). The benchmark falls in the Agentic category. Agents' Last Exam (Linux) is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About Agents' Last Exam (Linux)
Year
2026
Tasks
Linux agent tasks (subset size not published)
Format
Provider-reported task score
Difficulty
Advanced agentic work
Meta reports Muse Ember and its comparison models on the Linux subset in its internal evaluation suite, at max reasoning effort. Meta does not publish the subset size or version, so BenchLM keeps it apart from the full Agents' Last Exam key and the 86-task CLI subset and stores it display-only.
Freshness and provenance
Version
Agents' Last Exam (Linux) 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does Agents' Last Exam (Linux) measure?
The Linux subset of Agents' Last Exam, an end-to-end agent task benchmark, as reported by Meta.
Which model scores highest on Agents' Last Exam (Linux)?
Muse Ember by Meta currently leads with a score of 61.6% on Agents' Last Exam (Linux).
How many models are evaluated on Agents' Last Exam (Linux)?
3 AI models have published results on Agents' Last Exam (Linux) in the BenchLM catalog.
Compare top models on Agents' Last Exam (Linux)
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.