DeepSeek

DeepSeek V4.1 Flash 1017

DeepSeek's October update to V4.1 Flash. It scores higher on every test DeepSeek published, with shorter answers, at the same price.

deepseek/deepseek-v4.1-flash-1017Text and image inputtext + image inputtext output

Run this model

deepseek/deepseek-v4.1-flash-1017

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',    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 →

Use deepseek/deepseek-v4.1-flash-1017 to pin the October release. It accepts text and image input.

Text and image input with provider peak-hour pricingMiniRouter supports text and image input for this model. Prices shown include MiniRouter's markup. Final charges reflect Vercel's reported usage and any peak-hour surcharge. Source ↗

Model overview

Input / 1M
$0.07875
All-in MiniRouter rate
Output / 1M
$0.315
All-in MiniRouter rate
Context
1M
1,048,576 tokens
Max output
1M
tokens
DeepSWE v1.1
77.0%
DeepSeek-reported, up from 74.2%
Terminal-Bench 4.0
42.3%
DeepSeek-reported, up from 31.2%
Announced
17 Oct 2026
publisher release
Publisher evidence

Better agents for the same price.

DeepSeek compares the October release with September's V4.1 Flash, Claude Opus 5 and GPT-5.6 Sol in one table. Each chart has its own axis.

Terminal-Bench 4.0 ran in Claude Code and DeepSWE in Mini-SWE-V2. MiniRouter did not run these evals.

1017 vs September

DeepSWE v1.1

74.2% for September's V4.1 Flash rises to 77.0%.

DeepSeek-reported results

0 to 100%

  • V4.1 Flash 101777.0%
  • V4.1 Flash74.2%
  • Claude Opus 574.0%
  • GPT-5.6 Sol73.0%
DeepSeek V4.1 Flash 1017 announcement ↗, October 17, 2026. Publisher-reported.
View chart data
DeepSWE v1.1 values
ModelReported result
V4.1 Flash 101777.0%
V4.1 Flash74.2%
Claude Opus 574.0%
GPT-5.6 Sol73.0%

1017 vs September

Terminal-Bench 4.0

31.2% for September's V4.1 Flash rises to 42.3%.

DeepSeek-reported results

0 to 100%

  • V4.1 Flash 101742.3%
  • V4.1 Flash31.2%
  • Claude Opus 551.8%
  • GPT-5.6 Sol39.9%
DeepSeek V4.1 Flash 1017 announcement ↗, October 17, 2026. Publisher-reported.
View chart data
Terminal-Bench 4.0 values
ModelReported result
V4.1 Flash 101742.3%
V4.1 Flash31.2%
Claude Opus 551.8%
GPT-5.6 Sol39.9%

1017 vs September

ProgramBench

20.3% for September's V4.1 Flash rises to 29.1%.

DeepSeek-reported results

0 to 100%

  • V4.1 Flash 101729.1%
  • V4.1 Flash20.3%
  • Claude Opus 537.0%
  • GPT-5.6 Sol23.0%
DeepSeek V4.1 Flash 1017 announcement ↗, October 17, 2026. Publisher-reported.
View chart data
ProgramBench values
ModelReported result
V4.1 Flash 101729.1%
V4.1 Flash20.3%
Claude Opus 537.0%
GPT-5.6 Sol23.0%

1017 vs September

Automation-Bench V1.0.6

54.8% for September's V4.1 Flash rises to 60.5%.

DeepSeek-reported results

0 to 100%

  • V4.1 Flash 101760.5%
  • V4.1 Flash54.8%
  • Claude Opus 550.3%
  • GPT-5.6 Sol45.8%
DeepSeek V4.1 Flash 1017 announcement ↗, October 17, 2026. Publisher-reported.
View chart data
Automation-Bench V1.0.6 values
ModelReported result
V4.1 Flash 101760.5%
V4.1 Flash54.8%
Claude Opus 550.3%
GPT-5.6 Sol45.8%
Live route pricing

Plan around peak hours.

The price panel uses MiniRouter's catalog rates. DeepSeek charges 2× from 01:00–04:00 and 06:00–10:00 UTC, Monday–Friday, excluding Chinese public holidays.
  1. [ 01 ]

    Separate list prices from your route price

    DeepSeek lists $0.15/M input and $0.60/M output off-peak, and $0.30/M and $1.20/M at peak, before MiniRouter's markup.

  2. [ 02 ]

    Compare the whole task

    DeepSeek reports shorter outputs than September's release. Measure total task cost and success rate on your own prompts.

From V4.1 Flash

Pin the release you tested.

DeepSeek's API now serves this release as deepseek-flash. On MiniRouter, the dated ID keeps your requests on the October weights.
  1. [ 01 ]

    Change one ID

    Replace deepseek/deepseek-v4.1-flash with deepseek/deepseek-v4.1-flash-1017. Keep your MiniRouter base URL and key.

  2. [ 02 ]

    Rerun your evaluations

    Rerun representative tasks before switching production traffic. Publisher scores are a starting point, not a measurement of your workload.

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
Endpoints/v1/chat/completions/v1/messages/v1/responses
Parametersmax_tokens, temperature, stop, tools, tool_choice, reasoning, include_reasoning, response_format, structured_outputs
Structured outputJSON Schema and JSON mode enforced on verified providers.
ModalitiesInput: text, image · Output: text

Configured providers. Priority follows the live catalog order.

ProviderPriorityInput / 1MOutput / 1MParametersAccess
vercel
1$0.07875$0.315max_tokens, temperature, stop, tools, tool_choice, reasoning, include_reasoning, response_format, structured_outputsMiniRouter balance

Every model, one key.