Skip to content

[Feat] Add DeepSeek V4 Flash Vision Exp support - #1488

Merged
taltas merged 2 commits into
mainfrom
feature/deepseek-v4-flash-vision-1f1mkf2t5468c
Sep 5, 2026
Merged

[Feat] Add DeepSeek V4 Flash Vision Exp support#1488
taltas merged 2 commits into
mainfrom
feature/deepseek-v4-flash-vision-1f1mkf2t5468c

Conversation

@zoomote

@zoomote zoomote Bot commented Sep 1, 2026

Copy link
Copy Markdown
Contributor

​Opened on behalf of @taltas. Follow up by mentioning @roomote, in the web UI, or in Discord.

What changed

  • Adds deepseek-v4-flash-vision-exp to the native DeepSeek catalog with vision, thinking, temperature, context/output limits, cache pricing, and prompt-cache metadata.
  • Adds Fireworks' verified static entry, accounts/fireworks/models/deepseek-v4-flash-vision-exp, with provider-specific limits and pricing.
  • Extends native DeepSeek request handling for thinking controls and reasoning-effort normalization while preserving image blocks in OpenAI-compatible requests.
  • Adds focused coverage for metadata, model fetching, static selection, request behavior, and image input.

Why this change was made

DeepSeek released an experimental multimodal V4 Flash model that is now available through the native DeepSeek API and Fireworks. Static catalogs need explicit metadata so Zoo Code can select the model, retain image input, expose reasoning controls, and estimate cost correctly.

Authoritative sources:

Impact

Users can select the model through native DeepSeek and Fireworks with image support and accurate provider-specific metadata. Native peak rates are $0.44/M cache-miss input, $0.014/M cache-hit input, and $1.32/M output; DeepSeek's off-peak rates are half. Fireworks rates are $0.22/M input, $0.007/M cached input, and $0.66/M output.

OpenRouter already discovers this model dynamically, so no static entry was added. Other static provider catalogs were left unchanged because their providers do not currently expose a verified deployment and provider-specific ID. After merging current main, full lint and type checks pass, focused suites pass with 144 tests and 1 skipped, and the full Zoo Code suite passes with 8,006 tests and 39 skipped. The delegated visual-proof handoff timed out, so no screenshot evidence is included.

Related PRs

@zoomote zoomote Bot assigned taltas Sep 1, 2026
@github-actions

github-actions Bot commented Sep 1, 2026

Copy link
Copy Markdown
Contributor

Review status

This PR was opened by an automated account. A human maintainer must verify the change intent, provenance, and validation before merging.

Current step: The required review sequence passed. Remaining merge requirements apply.

Review-state labels are managed by this workflow; do not edit them manually.

@codecov

codecov Bot commented Sep 1, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

@github-actions github-actions Bot added has-conflicts PR has merge conflicts with the base branch and removed has-conflicts PR has merge conflicts with the base branch labels Sep 2, 2026
@github-actions github-actions Bot removed the has-conflicts PR has merge conflicts with the base branch label Sep 4, 2026
@taltas
taltas marked this pull request as ready for review September 4, 2026 23:15
@taltas
taltas added this pull request to the merge queue Sep 5, 2026
Merged via the queue into main with commit c4574ff Sep 5, 2026
35 of 36 checks passed
@taltas
taltas deleted the feature/deepseek-v4-flash-vision-1f1mkf2t5468c branch September 5, 2026 01:45
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants