Amux AI

DeepSeek: DeepSeek V4 Pro

Active

DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window. It is designed for advanced reasoning, coding, and long-horizon agent workflows, with strong performance across knowledge, math, and software engineering benchmarks.

By:
DeepSeek
Input:
Output:
Context length:
1M
Max output:
393K
Published:
2026-08-12
toolsreasoningstructuredOutputstreamingcaching

Providers

Same model, different providers. Automatic routing spreads requests across healthy providers by default, or ranks them by price, latency and more.

Provider
25% offdeepseekPeak$1.32$0.99/1M tokensOff-Peak$0.66$0.495/1M tokensPeak$3.96$2.97/1M tokensOff-Peak$1.98$1.485/1M tokens
ReadPeak$0.044$0.033/1M tokensOff-Peak$0.022$0.0165/1M tokensWrite—
1M78.24%2.2s100.47 tps
5% offalibaba-cloudOff-Peak$0.66$0.627/1M tokensPeak$1.32$1.254/1M tokensOff-Peak$1.98$1.881/1M tokensPeak$3.96$3.762/1M tokens
ReadOff-Peak$0.066$0.0627/1M tokensPeak$0.132$0.1254/1M tokensWrite—
1M94.13%2.6s100.52 tps
30% offLowestvolcenginePeak$1.32$0.924/1M tokensOff-Peak$0.66$0.462/1M tokensPeak$3.96$2.772/1M tokensOff-Peak$1.98$1.386/1M tokens
ReadPeak$0.044$0.0308/1M tokensOff-Peak$0.022$0.0154/1M tokensWrite—
1M90.42%9.6s7.05K tps
10% offtencent-cloudPeak$1.32$1.188/1M tokensOff-Peak$0.66$0.594/1M tokensPeak$3.96$3.564/1M tokensOff-Peak$1.98$1.782/1M tokens
ReadPeak$0.044$0.0396/1M tokensOff-Peak$0.022$0.0198/1M tokensWrite—
1M81.80%3.1s79.24 tps

Availability

24 hours · 10/21–10/22

Success rate of requests sent by the Amux AI gateway to each provider. Client errors are excluded.

Tok/s

24 hours · 10/21–10/22

Average output speed of streaming requests, in tokens per second (TPS).

TTFT

24 hours · 10/21–10/22

Average time to first token (TTFT) of streaming requests, measured from request dispatch to the first token received.

Activity

24 hours · 10/21–10/22

Relative request activity over time, broken down by serving provider. Indexed to the peak within the selected range (= 100); reflects trends and provider share, not absolute volume.

Related models

More models from DeepSeek

DeepSeek: DeepSeek V4.1 Flash1M contextfrom $0.105 / M tokens4 providersDeepSeek V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the first built on the company's Causal Encoder-Decoder (CED) architecture. It activates 8B parameters on input and 16B on output from a 552B-parameter backbone, an asymmetric split that keeps per-token compute low relative to the model's total size. Image understanding is native to the architecture, with visual and text embeddings trained jointly from the start of pre-training rather than added afterward as in the earlier experimental V4 Flash Vision Exp(opens in new tab).