BlogLLM Benchmark Scores as Published, Side by Side
APEX-SWE: Results as Published
Every result the catalog holds for APEX-SWE: 3 models from 2 developers, read from 1 page. Each number is as printed by the company named beside it; nothing is ranked or rescored. Models are listed by developer, newest first.
Area Agentic CodingMaintainer MercorIntroduced 2026Data version 2026103001, as of 2026-10-30
APEX-SWE tests whether an AI agent can do production software engineering work. Its 200 hidden tasks (plus a 50-task open development set) were written by working software engineers and split evenly between integration tasks, building end-to-end systems across cloud services, business applications and infrastructure-as-code, and observability tasks, diagnosing production failures from logs, dashboards and other unstructured context. Each task has validation tests and a gold-standard output, and the result is reported as Pass@1, the first-attempt success rate.
APEX-SWE
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| Claude Fable 5Anthropic, released 2026-06-09 | 58.8% | xAI, Introducing Grok 4.6checked 2026-10-15 | Effort: max |
| Grok 4.6xAI, released 2026-08-12 | 56.4% | xAI, Introducing Grok 4.6checked 2026-10-15 | Effort: high |
| Grok 4.5xAI, released 2026-07-16 | 53.6% | xAI, Introducing Grok 4.6checked 2026-10-15 | Effort: high |
Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.