DeepSeek V4 Flash
deepseek/deepseek-v4-flashVery low cost per token, with peak-hour pricing at 2x. Sourceable through Webparam on request; not provisioned on this account yet.
1.048576M ctx · R4.6 in · R13.7 out
深度求索DeepSeek
Open-weight frontier reasoning at disruptive economics
Models
5
Founded
2023
Country
China
Hangzhou
DeepSeek has been building models out of Hangzhou, China, since 2023. Webparam tracks five of its models, spanning language, reasoning and coding: priced from R0.00 to R13.70 per million input and reaching 1.048576M. Three of them ship with .
Popular models
Most used this week
The three highest-usage models from this lab on Webparam.
DeepSeek V4 Flash by DeepSeek DeepSeekOn request DeepSeek V4 Pro by DeepSeek DeepSeekOn request DeepSeek V4 Pro
deepseek/deepseek-v4-proFrontier-adjacent reasoning at a fraction of Western flagship cost. Sourceable through Webparam on request; not provisioned on this account yet.
1.048576M ctx · R13.7 in · R41.3 out
DeepSeek V3.2 by DeepSeek DeepSeekOn request DeepSeek V3.2
deepseek/deepseek-v3.2Frontier-adjacent open-weight flagship at roughly a tenth of typical flagship cost, with reliable tool calling.
128K ctx · Price on request
Full catalogue
All 5 models
Sorted by weekly usage. Choose any column header to reorder.
| Category | Output $/1M | Actions | ||||
|---|---|---|---|---|---|---|
DeepSeek V4 Flashdeepseek/deepseek-v4-flash | Language | 1.048576M | R4.60 | R13.70 | ||
DeepSeek V4 Prodeepseek/deepseek-v4-pro | Reasoning | 1.048576M | R13.70 | R41.30 | ||
DeepSeek V3.2deepseek/deepseek-v3.2 | Language | 128K | R0.00Free tier | — | ||
DeepSeek R1deepseek/deepseek-r1 | Reasoning | 164K | R0.00Free tier | — | ||
DeepSeek Coder V2deepseek/deepseek-coder-v2 | Coding | 128K | R0.00Free tier | — |
- DeepSeek V4 Flashdeepseek/deepseek-v4-flash1.048576M ctxR4.6 in · R13.7 out
- DeepSeek V4 Prodeepseek/deepseek-v4-pro1.048576M ctxR13.7 in · R41.3 out
- DeepSeek V3.2deepseek/deepseek-v3.2128K ctxPrice on request
- DeepSeek R1deepseek/deepseek-r1164K ctxPrice on request
- DeepSeek Coder V2deepseek/deepseek-coder-v2128K ctxPrice on request
Recent releases
What shipped lately
History
Spun out of a quantitative trading firm, DeepSeek built its reputation by releasing open-weight models that matched closed frontier quality at a fraction of the price. Its V-series redefined cost expectations for production LLMs, and its R-series brought open reasoning models into serious enterprise conversations.
Spun out of a quantitative trading firm, DeepSeek built its reputation by releasing open-weight models that matched closed frontier quality at a fraction of the price. Its V-series redefined cost expectations for production LLMs, and its R-series brought open reasoning models into serious enterprise conversations.
Strengths
- Frontier-adjacent quality at disruptive token prices
- Open weights across the flagship line
- Rapid, research-driven release cadence
Use DeepSeek models through Webparam
Every model above is callable through one OpenAI-compatible convention: change the model ID, keep the request.
Chat completions reference →DeepSeek V4 Flash model details →
Related providers