eval(qwen): add Qwen 3.8 quality benchmarks and native 14px pilot - #244

Merged

teamchong merged 11 commits into

mainfrom

eval/qwen-3.8-quality

Aug 21, 2026

Commits on Aug 19, 2026

eval(qwen): add Qwen 3.8 quality benchmarks and native 14px pilot
teamchong
committed
f0293e3

Commits on Aug 20, 2026

fix(openai): use 'high' detail for non-gpt5 models and address review findings
teamchong
committed
542d4e3
feat(qwen): add dedicated Qwen profile and dynamic image cap under 32 images
teamchong
committed
82ebbb0
fix(bridge): use detail high for non-gpt5 models in messages-responses-bridge
teamchong
committed
a60f337
fix(openai): count slab images once in the history-collapse budget
teamchong
committed
83c8911
fix(bridge): pick image detail from the request model
teamchong
committed
cf07f44
test(openai-gpt5): guard find() results instead of non-null assertions
teamchong
committed
4e9f97f
feat(profiles): key the Qwen profile to qwen3.8-27b exactly
teamchong
committed
1693de1
chore: drop private gateway references from code and docs
teamchong
committed
6b536d8
feat(qwen): budget history against the 32-image provider cap
teamchong
committed
73592d6

Commits on Aug 21, 2026

fix(qwen): prioritize history within image cap
teamchong
committed
081abcc