AFM 0.9.14 · release evidence

Qwen 3.6 release validation

PublishedJuly 17, 2026Qualified targetmlx-community/Qwen3.6-35B-A3B-4bit

The v0.9.14 validation record spanning Swift tests, comprehensive MLX coverage, assertions, Promptfoo agent behavior, and VulcanBench regression evidence.

Qualification results

Swift XCTest106 / 106

All XCTest cases passed.

Swift Testing373 / 373

All Swift Testing cases passed.

Comprehensive MLX190 / 190

Full model suite.

Assertion aggregate459 / 469

Each failed section passed on targeted rerun.

Promptfoo332 / 388

Compared with 325/388 for the v0.9.13 baseline.

VulcanBench375 passed

One skipped, one known failure, and two deselected.

Interpretation

Known caveats stay visible.

Ten assertion cases missed in the aggregate run; every affected section passed when rerun individually.
The single VulcanBench failure was the known Cargo/PATH error-ordering assertion.

Bundle inventory

152 evidence files.

Comprehensive MLX HTML and JSONL reports
Aggregate and targeted assertion reruns
v0.9.14 and v0.9.13-baseline Promptfoo results
Swift test logs and XML reports
VulcanBench XML report

Canonical evidence bundle

afm-v0.9.14-test-reports.tar.gz

1.60 MiB compressed archive attached to the AFM 0.9.14 GitHub Release.

Download reports ↗Open the GitHub Release ↗

SHA-256: a127e57afd223b7f5422241d1c3cb5cfd5a390625004b005670baaaa7d2e5dbb
Reference URL: https://maclocal.ai/test-reports/0.9.14