Last verified
IMAGE CAP 384 TOKENSSAME CARD AS V4 FLASHPEAK / OFF-PEAK BILLING1M CONTEXTEXPERIMENTAL ID

DeepSeek V4 Flash Vision (Exp) API Pricing

DeepSeek's first vision model costs exactly what the text model costs: at peak $0.44/M input, $1.32/M output and $0.014/M cached input, and half of each off-peak, on the same time-of-day card as V4 Flash. What makes images cheap here is not the rate but the ceiling: every image is resized to roughly 800x800 before inference, which caps it at 384 tokens — so a 5000x5000 photo and a 2000x2000 one bill the same, about $0.00017 at peak and half that off-peak. The -exp in the id is DeepSeek's own: this is an experimental endpoint, not a GA one, and it is the only DeepSeek model that accepts images at all.

Input - per 1M tokens
$0.44/M
Off-peak $0.22/M peak
Output - per 1M tokens
$1.32/M
Off-peak $0.66/M peak
Cached input
$0.014/M
Off-peak $0.007/M peak
Effective - agentic blend
$0.189/M
92/8 split - 82% cache
§ 01 / TERMINAL

Run the numbers.

Calculator pre-loaded with PEAK rates — halve every result for an off-peak run, which covers 17 of every 24 hours. Images enter the same input bucket at up to 384 tokens each, so an image-heavy workload is priced by counting images, not megabytes.

$ /mo
Workload split
Prompt cache hit rate
Tokens you can process
Words equivalent (English)
Effective rate
Open full calculator (all models · share URL · CSV) →
§ 02 / SCENARIOS

Real-world presets.

§ 03 / TOKENIZER

Paste text. See tokens. See cost.

Calibrated · measured on the vendor's tokenizer · 2026-06-10 Auto-counts as you type

Counts use a chars-per-token calibration measured on the vendor's own published tokenizer (deepseek-ai/DeepSeek-V3.2-Exp, 2026-06-10). English prose is typically within a few percent; code and non-Latin scripts tokenize heavier. For billing-exact counts use the vendor's count-tokens API.

Characters 497
Words 89
Tokens (estimated) 95 tokens
Cost as input · uncached $0.00004 USD
Cost as output · uncached $0.00013 USD
Cost as cached input $0.000001 USD
§ 04 / SHELF

Up against the shelf.

All models →
Model Input /M Output /M Effective blended Context Best for
DeepSeek V4 Flash Vision (Exp) $0.44 cache $0.014 $1.32 $0.189 current page - peak 1M The only DeepSeek that reads images
DeepSeek V4 Flash $0.44 cache $0.014 $1.32 $0.189 same 1M Identical card, text only
DeepSeek V4 Pro $1.32 cache $0.044 $3.96 $0.569 pricier 1M Harder reasoning, no image input
Qwen3-VL Flash $0.05 $0.40 $0.078 cheaper 256K Cheapest vision tier, shorter window
GPT-5.6 Luna $0.20 cache $0.02 $1.20 $0.152 cheaper 1.05M Cheaper vision on a GA endpoint
Gemini 3.7 Flash $0.75 cache $0.075 $3.75 $0.481 pricier 1M Vision plus audio and video input
§ 05 / DEEP LINKS

Specific scenarios.

All calculators →

Frequently asked.

What images actually cost on DeepSeek's experimental vision endpoint, and where the limits bite.

Q · 01 How much does one image cost? +
At most $0.00017 at peak and $0.000084 off-peak. Images are billed as ordinary input tokens, and DeepSeek resizes every image to roughly an 800x800 pixel budget before inference, which puts a hard ceiling of 384 tokens on any single image. A 2000x2000 screenshot and a 5000x5000 photo therefore cost the same. Roughly 5,900 images per dollar at peak, before the text tokens that go with them.
Q · 02 Does it cost more than the text model? +
No — the two share one rate card exactly: $0.44 / $0.014 / $1.32 at peak and half that off-peak. There is no vision surcharge, no separate image unit, and no minimum. The only reason to pick V4 Flash instead is that this endpoint is experimental.
Q · 03 What does the -exp in the model id mean? +
It is DeepSeek's label, not ours. The pricing page's MODEL VERSION row reads DeepSeek-V4-Flash-Vision-Exp, so the experiment is baked into the id you call. Treat it the way you would any experimental endpoint: the id can change, and DeepSeek publishes no deprecation window for it. We record it with preview status for that reason.
Q · 04 How many images can one request carry? +
Up to 600, subject to size limits: 32 MiB per image inline or by URL, 64 MiB via the Files API, 48 MiB for the whole request body, and a maximum dimension of 8192 px per side — which drops to 4096 px once a request holds 15 or more images. At the 384-token ceiling, 600 images come to about 230K tokens, or roughly $0.10 of input at peak.
Q · 05 What does it give up against V4 Flash? +
One documented capability: FIM completion. V4 Flash supports it in non-thinking mode; this model does not support it at all. Everything else on DeepSeek's feature table matches — JSON output, tool calls, the Responses API, the Anthropic-format endpoint, chat prefix completion, 1M context, 384K max output and a 2500 concurrency limit.
Q · 06 Are images allowed anywhere in the conversation? +
No — user messages only. An image in a system or assistant message returns a 400, and so does sending an image to any other DeepSeek model, which replies "This model does not support image". Supported formats are JPEG, PNG, GIF and WebP, detected from file content rather than from the file name.
Q · 07 Which hours are peak? +
01:00-04:00 and 06:00-10:00 UTC, Monday through Friday. Every other hour — 17 of 24 on weekdays, and all weekend — is off-peak at exactly half the rate. Batch-style image work that can wait is the clearest case for scheduling: the same 600-image request costs about $0.05 off-peak instead of $0.10.