Release evidence

Test reports, release by release.

Stable, linkable summaries of AFM qualification runs. The complete raw evidence remains attached to each GitHub Release as a digest-identified archive.

Cross-model release qualification

Final v0.9.19 evidence covering six native MLX target and MTP modes, DwarfStar DSpark, Context through 32K, llmprobe conformance, and installed Qwen 3.8 27B testing on M3 Ultra and M4 Pro.

Native MLX MUST1,166 / 1,166Context coverage7 × through 32KQwen comprehensive89 / 91 per host

AFM 0.9.18.1

A focused stable hotfix that restores the Apple Foundation Models provider in release builds and adds a hard gate rejecting artifacts that do not link FoundationModels.framework.

No permanent test-report bundle was attached to this release. Open the canonical GitHub release record for the published change history.

AFM 0.9.18

Completes the AFMKit provider-ownership migration, introduces the advanced native terminal chat UI, hardens streamed stop handling, and strengthens installed-package and evaluation-bundle verification.

No permanent test-report bundle was attached to this release. Open the canonical GitHub release record for the published change history.

AFM 0.9.17

Adds automatic, quantization-matched Qwen 3.8 MTP sidecar discovery and expands Qwen 3.8 tool-calling qualification.

No permanent test-report bundle was attached to this release. Open the canonical GitHub release record for the published change history.

Qwen 3.8 release qualification

Release-candidate qualification for Qwen 3.8 27B covering the Swift suite, live model behavior, CLI integration, comprehensive inference, and agentic Promptfoo scenarios.

Swift suite420 passedLive assertions118 / 118Comprehensive182 / 182

DeepSeek V4 release validation

Release evidence for the DeepSeek V4 AFM MLX path, including unit tests, live assertions, comprehensive inference, concurrency, prefix reuse, and Promptfoo agent workflows.

Swift suite390 passedAssertion rerun487 / 487Comprehensive182 / 182

Qwen 3.6 release validation

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

Swift XCTest106 / 106Swift Testing373 / 373Comprehensive MLX190 / 190