DeepSeek

DeepSeek V4.1 Flash 1017 Ultrafast

October's DeepSeek V4.1 Flash on Fireworks' Ultra-fast serving. Fireworks reports up to 1,500 tokens per second, at 8 times its standard token price.

deepseek/deepseek-v4.1-flash-1017-ultrafastPublic previewtext + image inputtext output

Run this model

deepseek/deepseek-v4.1-flash-1017-ultrafast

Use an endpoint supported by this model.

Get a key for this setup →
Thinking guide →
const response = await fetch('https://api.minirouter.sh/v1/chat/completions', {  method: 'POST',  headers: {    Authorization: `Bearer ${process.env.MINIROUTER_KEY}`,    'Content-Type': 'application/json',  },  body: JSON.stringify({    model: 'deepseek/deepseek-v4.1-flash-1017-ultrafast',    messages: [{ role: 'user', content: 'Why is the sky blue?' }],    max_tokens: 1024,  }),}) const data = await response.json()console.log(data.choices[0].message.content)
Read docs →

Same weights as deepseek/deepseek-v4.1-flash-1017, served faster. Set max_tokens explicitly; the endpoint is a preview and can change.

Model overview

Input / 1M
$2.52
All-in MiniRouter rate
Output / 1M
$10.08
All-in MiniRouter rate
Context
1M
1,048,576 tokens
Max output
1M
tokens
Speed
Up to 1,500 tok/s
Fireworks-reported
Price
8× standard
$2.40 / $9.60 Fireworks list
Released
6 Nov 2026
publisher release

Compare text-model alternatives or estimate token costs for your workload.

Publisher-reported, plotted by MiniRouter

A one-minute agent task in eight seconds.

Fireworks timed a 30-turn agent task on standard and Ultra-fast serving. It did not publish the task or hardware setup, and MiniRouter did not run this test.

30-turn agent task

Completion time, seconds

0 to 60 seconds

  • Ultra-fast8
  • Standard60

Shorter is better.

Fireworks list, per 1M tokens

Input
$2.40standard $0.30
Cached input
$0.048standard $0.006
Output
$9.60standard $1.20
Fireworks announcement ↗, November 3, 2026, and serverless pricing ↗. Prices before MiniRouter's fee.
View chart data
Fireworks standard and Ultra-fast serving for DeepSeek V4.1 Flash 1017
MeasureUltra-fastStandard
30-turn agent task8 seconds60 seconds
Input / 1M$2.40$0.30
Cached input / 1M$0.048$0.006
Output / 1M$9.60$1.20
Three ways to run 1017

Pay for speed where waiting costs more.

All three IDs serve the October weights. Fireworks says Ultra-fast changes generation speed, not the model.
  1. [ 01 ]

    Ultrafast for interactive loops

    Use it when a person or a long agent run waits on every turn. Compare completed-task time and cost with the standard model.

  2. [ 02 ]

    Standard for volume

    deepseek/deepseek-v4.1-flash-1017 costs far less per token and routes across several providers. Keep batch and background work there.

  3. [ 03 ]

    Expect preview changes

    Fireworks runs Ultra-fast as a public preview with one provider route. Keep a fallback model for production traffic.

Route

One route, standard clients.

Authenticate with a MiniRouter bearer key. Caller-provided upstream credentials are not used on this route.
Base URLhttps://api.minirouter.sh/v1
Model IDdeepseek/deepseek-v4.1-flash-1017-ultrafast
Endpoints/v1/chat/completions/v1/messages/v1/responses
Parametersmax_tokens, temperature, stop, tools, tool_choice, reasoning, include_reasoning
Structured outputNot enforced; a requested format is dropped and the answer is plain text.
ModalitiesInput: text, image · Output: text

Configured providers. Priority follows the live catalog order.

ProviderPriorityInput / 1MOutput / 1MParametersAccess
vercel
1$2.52$10.08max_tokens, temperature, stop, tools, tool_choice, reasoning, include_reasoningMiniRouter balance

Every model, one key.