deepseek-v4-flash API — compare providers

The same model (deepseek-v4-flash) is served by 22 providers; Baidu is the cheapest at $0.15/1M blended — the priciest (Cloudflare) costs 1073% more. On quality, deepseek-v4-flash scores 1436 on the LMArena (Chatbot Arena) — so the cheapest endpoint is also the best value per dollar here.

LicenseMITopen weights · permissive · terms
Knowledge cutoffMarch 2026source · ranked
Quality (Elo)1436LMArena (Chatbot Arena) · source
Intelligence51.8Artificial Analysis Intelligence Index · source · ranked

Quality and intelligence scores are attached to the model itself — identical whichever provider serves it. Verified 2026-09-04.

RankProviderInput /1MOutput /1MBlended /1MValue (Elo/$)
1Baidu$0.05$0.1$0.15 ← cheapest9,573 ← best value
2Relace$0.065$0.18$0.2455,861
3AkashML$0.065$0.18$0.2455,861
4DeepInfra$0.08$0.18$0.265,523
5DigitalOcean$0.08$0.252$0.3324,325
6Morph$0.0987$0.278$0.3773,812
7BaseTen$0.13$0.26$0.393,682
8CoreWeave$0.13$0.28$0.413,502
9Together AI$0.14$0.28$0.423,419
10Parasail$0.14$0.28$0.423,419
11Venice$0.175$0.35$0.5252,735
12Alibaba$0.176$0.528$0.7042,040
13Reka$0.11$0.66$0.771,865
14Azure OpenAI$0.21$0.56$0.771,865
15Fireworks AI$0.22$0.66$0.881,632
16SiliconFlow$0.22$0.66$0.881,632
17DeepSeek$0.22$0.66$0.881,632
18GMICloud$0.352$1.056$1.4081,020
19Novita AI$0.4092$1.2276$1.637877
20Phala$0.44$1.32$1.76816
21AtlasCloud$0.44$1.32$1.76816
22Cloudflare$0.44$1.32$1.76816

Value = model Elo ÷ blended price per 1M tokens. Since the model's quality is identical across providers, the cheapest endpoint is also the best value per dollar.

Prices are per 1M tokens (USD). "Blended" = input + output, for coarse ranking. Same underlying model, 22 serving providers, 1073% spread top to bottom. Context window: 1,310,720 tokens (131,072 max output) — see how it ranks in biggest context windows. See how this compares across the catalogue in same model, different price, or find the best quality per dollar in the value ranking.

Cost per 1,000 requests by workload

What each provider actually bills for a representative job, not just the sticker price.

ProviderChatbot
1000 in / 500 out
RAG / long context
8000 in / 500 out
Batch summarize
4000 in / 1000 out
Baidu$0.10$0.45$0.30
Relace$0.15$0.61$0.44
AkashML$0.15$0.61$0.44
DeepInfra$0.17$0.73$0.50
DigitalOcean$0.21$0.77$0.57
Morph$0.24$0.93$0.67
BaseTen$0.26$1.17$0.78
CoreWeave$0.27$1.18$0.80
Together AI$0.28$1.26$0.84
Parasail$0.28$1.26$0.84
Venice$0.35$1.57$1.05
Alibaba$0.44$1.67$1.23
Reka$0.44$1.21$1.10
Azure OpenAI$0.49$1.96$1.40
Fireworks AI$0.55$2.09$1.54
SiliconFlow$0.55$2.09$1.54
DeepSeek$0.55$2.09$1.54
GMICloud$0.88$3.34$2.46
Novita AI$1.02$3.89$2.86
Phala$1.10$4.18$3.08
AtlasCloud$1.10$4.18$3.08
Cloudflare$1.10$4.18$3.08

Estimate your own workload

Input tokens/request:   Output tokens/request:   Requests:

ProviderEstimated cost (USD)

Self-host / quantized builds

Open weights under MIT — you can run deepseek-v4-flash on your own hardware instead of paying per token. Quantized builds lower the memory footprint; whether it fits your GPU or Mac depends on the model's size. Formats: GGUF for llama.cpp / Ollama / LM Studio, AWQ and GPTQ for GPU serving, MLX for Apple silicon.

Links search the live model hubs (Hugging Face, Ollama) so they track new builds as they appear — we don't host weights. Which formats exist depends on what the model owner and community have published.