youssofal/MTPLX
The fastest way to run Qwen 3.8 Flash Next, Qwen 3.8 27B and Ternary Bonsai 2 27B on a Mac: 125 tok/s in OpenCode on an M5 Max, and a 27B model on 16 GB Macs. Native MTP speculative decoding on Apple Silicon, exact at any temperature. OpenAI and Anthropic compatible local server.
- Repository: https://github.com/youssofal/MTPLX
- Commit:
eb580fce3fe4465bafb261420e6fed794515dd61 - Generated: 2026-11-14T00:18:26+00:00
- Scorer 0.1.0, scoring prov-4+qual-2+sec-2, report schema v2
Scores
| Pillar | Score | Signal | Confidence | Coverage |
|---|---|---|---|---|
| AI provenance (estimated AI-generated share) | 3 | explicit | 95% | 100% |
| Code quality | 68 | adequate | 90% | 100% |
| Security | 54 | weak | 85% | 100% |
AI provenance
Explicit AI attribution on 3% of first-party code added; an estimated 3% including velocity signals.
- Explicitly attributed share (lower bound): 2.6%
- Unattributed share in AI-consistent velocity bursts: 46.6%
- Estimated AI-generated share: 2.6%
| AI tool | Attributed commits | Lines added |
|---|---|---|
| OpenAI Codex | 61 | 17,099 |
| Claude Code | 8 | 1,914 |
- AI tooling configured in repo: claude-code
Current source lines last changed by AI-attributed commits: 11,134 of 399,994 (sampled).
| Findings per 1,000 lines | AI-attributed | Other | Ratio |
|---|---|---|---|
| Quality | 0.629 | 0.494 | 1.27x |
| Security | 0.898 | 0.445 | 2.02x |
- Stylometry index: 0.01 per 1,000 lines (narrated steps 1, placeholders 0, assistant phrasing 1, emoji 3). Reported as a signal, not scored.
| Month | Lines added | Attributed to AI |
|---|---|---|
| 2026-05 | 95,479 | 12% |
| 2026-06 | 167,260 | 0% |
| 2026-07 | 80,186 | 2% |
| 2026-08 | 181,556 | 1% |
| 2026-09 | 189,538 | 2% |
| 2026-10 | 16,817 | 0% |
| 2026-11 | 4,107 | 0% |
Evidence
- OpenAI Codex: on merged branch (406b5f768e) (dba84f4d5bdd)
- OpenAI Codex: on merged branch (406b5f768e) (d2d95c4cb0f1)
- OpenAI Codex: on merged branch (406b5f768e) (b7d357d92974)
- OpenAI Codex: on merged branch (406b5f768e) (8d7f71e8be83)
- OpenAI Codex: on merged branch (406b5f768e) (7c9acd6c01e3)
- OpenAI Codex: on merged branch (406b5f768e) (906161901391)
- OpenAI Codex: on merged branch (406b5f768e) (6d76af544eff)
- OpenAI Codex: on merged branch (406b5f768e) (c0240b58269f)
- OpenAI Codex: on merged branch (406b5f768e) (1190dad5e18c)
- OpenAI Codex: on merged branch (406b5f768e) (306b24e76cf7)
- OpenAI Codex: on merged branch (406b5f768e) (096404b21dd9)
- OpenAI Codex: on merged branch (406b5f768e) (1ae398a1f246)
- OpenAI Codex: on merged branch (406b5f768e) (452b2e4ff95b)
- OpenAI Codex: on merged branch (406b5f768e) (dd92795339c3)
- OpenAI Codex: on merged branch (406b5f768e) (aee28241ffc0)
AI provenance describes how code was produced; it is not a quality judgment. Unless attribution is explicit, these are signals consistent with AI generation, not proof. Squash/rebase workflows, imported code, and rewritten history reduce accuracy.
Code quality
Code quality is adequate: test-to-source ratio 0.59, CI runs tests; 31% of function code in complex or long functions; 10% duplicated lines. Weakest area: type safety.
| Area | Score |
|---|---|
| Complexity (lizard) | 40 |
| Duplication | 55 |
| Code smells (Opengrep) | 81 |
| Tests and CI | 90 |
| Documentation | 100 |
| Type safety | 1 |
| Process (Scorecard review/maintenance) | n/a |
- Functions: 19,528; mean complexity 4.3, 90th percentile 9; 769 complex and 299 long functions
- Duplicated lines: 10.0% of 317,407 compared
- Tests: 237,537 test lines for 401,655 source lines (ratio 0.59); CI test commands: pytest; 86% of recent source commits also change tests
- Docs missing or thin: none
Most complex functions:
- generate_mtpk in mtplx/generation.py: complexity 990, 4837 lines
- cmd_serve_public in mtplx/commands/public.py: complexity 188, 758 lines
- b8 in mtplx/dashboard/_static/assets/index-D8G5hadf.js: complexity 173, 1 lines
- (anonymous) in mtplx/dashboard/_static/assets/index-D8G5hadf.js: complexity 161, 1 lines
- GM in mtplx/dashboard/_static/assets/index-D8G5hadf.js: complexity 150, 1 lines
Security
Security posture is weak: 69 known vulnerabilities in 10 dependencies; 85 high-severity SAST findings; 15 high-severity workflow issues. Weakest area: ci actions.
| Area | Score |
|---|---|
| Secrets (gitleaks, tree + history) | 65 |
| Static analysis (Opengrep) | 92 |
| Vulnerable dependencies (OSV) | 29 |
| CI workflows (zizmor) | 4 |
| Posture (policy, updates, signing, lockfiles) | 80 |
| OpenSSF Scorecard | n/a |
-
Known vulnerabilities: 69 in 10 packages (critical 6, high 35, medium 23, low 4)
-
Posture checks passed: security policy, dependency updates, lockfiles; missing: signed commits
Top findings
| Severity | Finding | Location |
|---|---|---|
| critical | osv/GHSA-5x8w-8jj5-q52h: [email protected] (PyPI) is affected by PYSEC-2026-4330, GHSA-5x8w-8jj5-q52h |
uv.lock |
| critical | osv/GHSA-8g5g-8vmr-9782: [email protected] (PyPI) is affected by PYSEC-2026-4332, GHSA-8g5g-8vmr-9782 |
uv.lock |
| critical | osv/GHSA-v9pv-5gc8-w2m6: [email protected] (PyPI) is affected by PYSEC-2026-4339, GHSA-v9pv-5gc8-w2m6 |
uv.lock |
| critical | osv/GHSA-xvh5-g6gv-559v: [email protected] (PyPI) is affected by PYSEC-2026-4363, GHSA-xvh5-g6gv-559v |
uv.lock |
| critical | osv/GHSA-332p-7hrc-92wf: [email protected] (PyPI) is affected by GHSA-332p-7hrc-92wf |
uv.lock |
| critical | osv/GHSA-w889-685w-c8fh: [email protected] (PyPI) is affected by GHSA-w889-685w-c8fh |
uv.lock |
| high | opengrep/javascript.browser.security.insecure-document-method: User controlled data in methods like innerHTML, outerHTML or document.write is an anti-pattern that can lead to XSS vulnerabilities [confidence low] |
dashboard/src/lib/uplotTooltip.ts:87 |
| high | opengrep/javascript.browser.security.insecure-innerhtml: User controlled data in a box.innerHTML is an anti-pattern that can lead to XSS vulnerabilities [confidence low] |
dashboard/src/lib/uplotTooltip.ts:87 |
| high | opengrep/python.lang.security.audit.dangerous-subprocess-use-audit: Detected subprocess function 'Popen' without a static string. If this data can be controlled by a malicious actor, it may be an instance of command injection. Audit the use of this call to ensure it is not controllable by an external resource. You may consider using 'shlex.escape()'. [confidence low] |
mtplx/aria2_downloader.py:234 |
| high | opengrep/python.lang.security.audit.dangerous-subprocess-use-audit: Detected subprocess function 'run' without a static string. If this data can be controlled by a malicious actor, it may be an instance of command injection. Audit the use of this call to ensure it is not controllable by an external resource. You may consider using 'shlex.escape()'. [confidence low] |
mtplx/benchmarks/code_eval.py:293 |
| high | opengrep/python.lang.security.audit.dangerous-subprocess-use-audit: Detected subprocess function 'run' without a static string. If this data can be controlled by a malicious actor, it may be an instance of command injection. Audit the use of this call to ensure it is not controllable by an external resource. You may consider using 'shlex.escape()'. [confidence low] |
mtplx/benchmarks/runners/preflight.py:36 |
| high | opengrep/python.sqlalchemy.security.sqlalchemy-execute-raw-query: Avoiding SQL string concatenation: untrusted input concatenated with raw SQL query can result in SQL Injection. In order to execute raw query safely, prepared statement should be used. SQLAlchemy provides TextualSQL to easily used prepared statement with named parameters. For complex SQL composition [confidence low] |
mtplx/cache_bank/cold_tier.py:3108 |
| high | opengrep/python.sqlalchemy.security.sqlalchemy-execute-raw-query: Avoiding SQL string concatenation: untrusted input concatenated with raw SQL query can result in SQL Injection. In order to execute raw query safely, prepared statement should be used. SQLAlchemy provides TextualSQL to easily used prepared statement with named parameters. For complex SQL composition [confidence low] |
mtplx/cache_bank/cold_tier.py:3111 |
| high | opengrep/python.lang.security.audit.dangerous-subprocess-use-audit: Detected subprocess function 'Popen' without a static string. If this data can be controlled by a malicious actor, it may be an instance of command injection. Audit the use of this call to ensure it is not controllable by an external resource. You may consider using 'shlex.escape()'. [confidence low] |
mtplx/commands/forge.py:1490 |
| high | opengrep/python.lang.security.audit.dangerous-subprocess-use-audit: Detected subprocess function 'run' without a static string. If this data can be controlled by a malicious actor, it may be an instance of command injection. Audit the use of this call to ensure it is not controllable by an external resource. You may consider using 'shlex.escape()'. [confidence low] |
mtplx/commands/forge.py:2347 |
| high | opengrep/python.lang.security.audit.dangerous-subprocess-use-audit: Detected subprocess function 'Popen' without a static string. If this data can be controlled by a malicious actor, it may be an instance of command injection. Audit the use of this call to ensure it is not controllable by an external resource. You may consider using 'shlex.escape()'. [confidence low] |
mtplx/commands/forge.py:3017 |
| high | opengrep/python.lang.security.audit.dangerous-subprocess-use-audit: Detected subprocess function 'Popen' without a static string. If this data can be controlled by a malicious actor, it may be an instance of command injection. Audit the use of this call to ensure it is not controllable by an external resource. You may consider using 'shlex.escape()'. [confidence low] |
mtplx/commands/forge.py:3119 |
| high | opengrep/python.lang.security.audit.dangerous-subprocess-use-audit: Detected subprocess function 'run' without a static string. If this data can be controlled by a malicious actor, it may be an instance of command injection. Audit the use of this call to ensure it is not controllable by an external resource. You may consider using 'shlex.escape()'. [confidence low] |
mtplx/commands/public.py:4839 |
| high | opengrep/python.lang.security.audit.dangerous-subprocess-use-audit: Detected subprocess function 'run' without a static string. If this data can be controlled by a malicious actor, it may be an instance of command injection. Audit the use of this call to ensure it is not controllable by an external resource. You may consider using 'shlex.escape()'. [confidence low] |
mtplx/commands/public.py:4921 |
| high | opengrep/python.lang.security.audit.dangerous-subprocess-use-audit: Detected subprocess function 'run' without a static string. If this data can be controlled by a malicious actor, it may be an instance of command injection. Audit the use of this call to ensure it is not controllable by an external resource. You may consider using 'shlex.escape()'. [confidence low] |
mtplx/commands/public.py:5208 |
| high | opengrep/python.lang.security.audit.dangerous-subprocess-use-audit: Detected subprocess function 'run' without a static string. If this data can be controlled by a malicious actor, it may be an instance of command injection. Audit the use of this call to ensure it is not controllable by an external resource. You may consider using 'shlex.escape()'. [confidence low] |
mtplx/commands/public.py:6807 |
| high | opengrep/python.lang.security.audit.dangerous-subprocess-use-audit: Detected subprocess function 'run' without a static string. If this data can be controlled by a malicious actor, it may be an instance of command injection. Audit the use of this call to ensure it is not controllable by an external resource. You may consider using 'shlex.escape()'. [confidence low] |
mtplx/commands/public.py:7872 |
| high | opengrep/python.lang.security.audit.dangerous-subprocess-use-audit: Detected subprocess function 'run' without a static string. If this data can be controlled by a malicious actor, it may be an instance of command injection. Audit the use of this call to ensure it is not controllable by an external resource. You may consider using 'shlex.escape()'. [confidence low] |
mtplx/commands/public.py:8332 |
| high | opengrep/python.lang.security.audit.dangerous-subprocess-use-audit: Detected subprocess function 'run' without a static string. If this data can be controlled by a malicious actor, it may be an instance of command injection. Audit the use of this call to ensure it is not controllable by an external resource. You may consider using 'shlex.escape()'. [confidence low] |
mtplx/commands/public.py:8518 |
| high | opengrep/python.lang.security.audit.dangerous-subprocess-use-audit: Detected subprocess function 'call' without a static string. If this data can be controlled by a malicious actor, it may be an instance of command injection. Audit the use of this call to ensure it is not controllable by an external resource. You may consider using 'shlex.escape()'. [confidence low] |
mtplx/commands/public.py:8569 |
Static signals only: the scorer never runs the project's code. Secret findings are pattern matches that were not verified, and secret values are never stored.
Distribution signals
- binary-release-asset: release v2.12.3 ships MTPLX-2.12.3.dmg (2572 downloads)
- binary-release-asset: release v2.12.3 ships mtplx-2.12.3.tar.gz (1783 downloads)
- binary-release-asset: release v2.12.2 ships MTPLX-2.12.2.dmg (8963 downloads)
- binary-release-asset: release v2.12.2 ships mtplx-2.12.2.tar.gz (4474 downloads)
- binary-release-asset: release v2.12.1 ships MTPLX-2.12.1.dmg (192 downloads)
Signals consistent with a repository that exists to distribute a download. They are heuristics, not a verdict; release files were not downloaded or inspected.
Repository profile
- Source lines: 401,655; test lines: 237,537
| Language | Lines |
|---|---|
| Python | 511,233 |
| Swift | 114,948 |
| TypeScript | 6,221 |
| Shell | 3,214 |
| C++ | 2,672 |
| C | 618 |
| JavaScript | 286 |