Skip to content

fix(core): include 'low' effort for deepseek-v4-flash #3732

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
yunaremaia wants to merge 2 commits into
base: main
Choose a base branch
from
+7 −3
Open

fix(core): include 'low' effort for deepseek-v4-flash #3732

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4efab85
fix(core): include 'low' effort for deepseek-v4-flash
yunaremaia Aug 24, 2026
b74a137
test(runtime): cover 'low' effort for deepseek-v4-flash
yunaremaia Aug 24, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 2 additions & 1 deletion packages/core/src/model-metadata.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -432,7 +432,8 @@ const STATIC_MODEL_METADATA: Partial<Record<ProviderType, Record<string, ModelMe
deepseek: {
'deepseek-v4-flash': {
capabilities: { ...REASONING_FUNCTION_CALLING, webSearch: true },
thinkingOptions: { efforts: ['high', 'max'], toggle: true },
lastUpdated: '2026-08-24',
thinkingOptions: { efforts: ['low', 'high', 'max'], toggle: true },
},
'deepseek-v4-pro': {
capabilities: { ...REASONING_FUNCTION_CALLING, webSearch: true },
Expand Down
7 changes: 5 additions & 2 deletions packages/runtime/src/__tests__/model-factory-thinking.test.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -301,18 +301,21 @@ describe('buildProviderOptions: thinking level', () => {
);
assert.deepEqual(
[...thinkingVariantsForModel('deepseek', 'deepseek-v4-flash')],
['high', 'max'],
['low', 'high', 'max'],
);
// DeepSeek V4 uses the generic Open Responses adapter, which passes a
// provider-native reasoningEffort through verbatim: `max` stays `max`
// (DeepSeek's documented mapping sends `xhigh` to high, not max).
assert.deepEqual(buildProviderOptions(conn('deepseek'), 'deepseek-v4-flash', 'low'), {
deepseek: { reasoningEffort: 'low' },
});
assert.deepEqual(buildProviderOptions(conn('deepseek'), 'deepseek-v4-flash', 'high'), {
deepseek: { reasoningEffort: 'high' },
});
assert.deepEqual(buildProviderOptions(conn('deepseek'), 'deepseek-v4-flash', 'max'), {
deepseek: { reasoningEffort: 'max' },
});
for (const unsupported of ['off', 'low', 'medium', 'minimal'] as const) {
for (const unsupported of ['off', 'medium', 'minimal'] as const) {
assert.deepEqual(
buildProviderOptions(conn('deepseek'), 'deepseek-v4-flash', unsupported),
{},
Expand Down