Tasks and inputs
DeepSeek V4 Flash Vision Experimental is listed for Image-text-to-text, Text generation. It accepts text, image and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
SyncDev model directory · Profile updated Oct 13, 2026
DeepSeek V4 Flash Vision Experimental adds image understanding to the V4 Flash architecture for visual reasoning and agents.
The specification score summarizes recorded features and limits. It is not a benchmark result.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
DeepSeek V4 Flash Vision Experimental adds image understanding to the V4 Flash architecture for visual reasoning and agents.
This release extends DeepSeek V4 Flash with a visual encoder and alignment components. It accepts images alongside text and returns text. The publisher presents it as an experimental multimodal checkpoint; it should be evaluated separately from a text-only V4 Flash API route.
The repository separates prompt encoding from reference inference and supplies both JSON content blocks and compact image-path notation. The released configuration sets 1,048,576 positions. Supported vLLM and SGLang recipes describe the vision path, tool parser, reasoning parser and DSpark configuration. Use the recipe for this exact checkpoint.
Give the agent a screenshot that changes the interpretation of a code issue, a chart question and a task with an absent visual detail. Check whether the final answer uses the supplied image correctly and whether tool arguments reflect that evidence. Keep the same screenshots and task harness across comparisons.
The complete weight set is roughly 305B parameters and uses MIT licensing. Sparse computation and speculative decoding do not remove the memory needed to serve the checkpoint. This profile leaves hosted prices unknown and does not assume that an existing DeepSeek API account exposes the experimental vision model. Confirm the available route before configuring a managed client.
The original first-party V4 Flash and V4 Flash Vision Exp models have retired. Their legacy API names are still accepted but are served by V4.1 Flash at its rates. Use deepseek-flash to request the current model explicitly. Downloadable checkpoints and independent hosts have their own deployment lifecycle.
Compare the model on the work you need to complete.
DeepSeek V4 Flash Vision Experimental is listed for Image-text-to-text, Text generation. It accepts text, image and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
Documented limits: 1.05M context. Measure a completed task using the prompts, documents and tools your application will use. Include reasoning and retries in the result.
The weights are downloadable under MIT. Check runtime support and size the complete model and working context for your expected concurrency.
These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.
Anthropic
Claude’s small model for high-volume extraction, classification and routing, with adaptive thinking and a 1M-token context window.
Mistral AI
Mistral Large 4 is an open-weight multimodal model released in public preview on the publisher’s announced release, for general reasoning, coding and agent applications.
OpenAI
GPT-6.1 Sol is OpenAI's updated Sol model for coding and professional tasks. It accepts text and images and returns text.
inclusionAI
Ling 3.1 Flash is inclusionAI’s hybrid reasoning model for coding and tool agents, available through a hosted AI Gateway route.
Anthropic
Claude Sonnet 5.5 is an Anthropic language model that accepts text and images and returns text.
MiniMax
MiniMax M3.1 Flash Preview is a multimodal reasoning model for code and agent workflows, accessed through eligible MiniMax plans and tools.