← All models
deepseek
API MODEL PROFILE / deepseek

DeepSeek: DeepSeek V3.2 Exp

deepseek/deepseek-v3.2-exp
Run model
Catalogue observation · 2026-09-22T10:10:29.140Z · Source: OpenRouter
textText API available

About this API

DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism...

Provider-supplied description, not an independently validated performance claim.
Input / 1 million tokens$0.27Published prompt tariff
Output / 1 million tokens$0.41Published completion tariff
Context window163,840Published token capacity

These are the catalogue’s token tariffs. Image, audio, video, search and other charges may use different billing units. Zero token rates alone do not mean those services are free. See the complete pricing fields below and the source model page.

Pros and useful characteristics
  • Published input token pricing is at or below the median of catalogue entries with the same output modalities and known tariffs.
  • Published output token pricing is at or below the same-modality catalogue median.
  • The published 163,840-token context window can accommodate larger inputs, subject to gateway limits.
  • No separate per-request fee is listed in this catalogue observation. Other modality charges may apply.
  • The upstream catalogue advertises tool-call parameters. This gateway does not expose them yet.
Cons and practical limits
  • Trident Protocol currently limits requests to text content, 24 KB of conversation and at most 4,096 output tokens. Free variants remain subject to upstream rate limits and availability.
  • No independent latency, uptime, reasoning-quality or coding benchmark has been measured by this application.

Assessment uses published fields and unweighted price medians across 426 entries with the same output modalities and known tariffs. It is a cost-and-capability comparison, not a quality ranking or recommendation.

Observed price history

No observations yet. The chart starts with the first recorded data point.

No observations yet. The chart starts with the first recorded data point.

One catalogue observation per hour. History begins when this deployment first observes the model. No backfilled or interpolated prices.

Published model tariffsUSD / 1M tokens
InputOutput
DeepSeek: DeepSeek V3.2 Exp
$0.27
$0.41

Zero baseline. Published token tariffs, not total multimodal cost or performance scores.

Estimate a request

Published tariff arithmetic; not a guaranteed bill or a gateway reservation.

$0.000475 estimated USD

Excludes unlisted provider-specific charges, taxes and any third-party markup. Context limits, reasoning tokens, caching and provider routing can affect the actual result.

Technical profile
Canonical model
deepseek/deepseek-v3.2-exp
Separate request tariff
$0
Upstream output limit
65,536
Gateway output limit
4,096
Upstream inputs
text
Upstream outputs
text
Knowledge cutoff
2025-07-31

Advertised upstream parameters

frequency_penaltyinclude_reasoninglogit_biaslogprobsmax_tokensmin_ppresence_penaltyreasoningrepetition_penaltyresponse_formatseedstopstructured_outputstemperaturetool_choicetoolstop_ktop_logprobstop_p

Upstream capabilities are separate from the text subset exposed by this gateway.

View source model page

Complete published pricing

Source fields are retained exactly, including extra charges and tier overrides. Negative values indicate variable pricing, not rebates. An absent field is not evidence that a service is free.

View OpenRouter pricing fields
{
  "prompt": "0.00000027",
  "completion": "0.00000041"
}
View deepseek profile ↗

Use this model with TRIDENT

Connect your wallet, verify your TRIDENT holdings and create an API key. Requests draw from the shared OpenRouter treasury.

Read the API access guide