Alexander Whitestone
|
adcb5b1ea9
|
feat: evaluate Qwen3.5:35B as local model option (#288)
Forge CI / smoke-and-build (pull_request) Failing after 48s
Part of Epic #281 -- Vitalik's Secure LLM Architecture.
Evaluation of Qwen3.5-35B-A3B (MoE, 35B total / 3B active) for local
deployment as privacy-sensitive inference tier.
- scripts/evaluate_qwen35.py: specs, VRAM, hardware matrix, security
scoring (Vitalik framework 8.8/10), fleet comparison, integration
- tests/test_evaluate_qwen35.py: 9 tests
Verdict: APPROVED. Perfect data locality, 128K context, Apache 2.0,
MoE speed advantage, tool use supported, eliminates Privacy Filter.
Closes #288
|
2026-04-13 21:13:17 -04:00 |
|