DeepSeek V4.1 Flash

DeepSeek · 1 offer

Free to try
cheapest in, per M
$0.30
cheapest out, per M
$1.20
context, tokens
1,048,576
typical first answer
12.7 s
uptime, 7 days
40.7%

Measured

From OpenVibe.Report: real calls in the last 30 days, with no prompt and no person in them.

Not measured yet: a model is measured after 20 calls in 30 days.

Every offer that serves it

Price in, out and cached per million tokens; speed and uptime from the last seven days of real requests; the privacy label and the provider’s own data policy.

OfferStatusIn / out per MSpeedFirst tokenUptimeErrorsPrivacyData policyTry
DeepSeek Free to tryUp$0.30 / $1.2012.7 s p50 · 26.2 s p95–40.7%59.3%Anonymized
refuses 0.0% (measured 2026-11-07)		not statedTry it

Compare DeepSeek V4.1 Flash with another model →

Activity

Requests a day across every offer.

  1. 072026-11-07: 1 requests
  2. 102026-11-10: 3 requests
  3. 112026-11-11: 2 requests
  4. 122026-11-12: 2 requests
  5. 132026-11-13: 3 requests

Measured performance

No completed day has twenty qualifying runs for this model.

Metrics and rating data

Bench rating: chat.arena

No single model rating has twenty qualifying games.

Source: OpenVibe.Bench. Latency and cost come from OpenVibe.AI receipts.

Presets

A preset fixes the router’s floor, ceiling and policy, and you name it as the model.

Cheapestmodel: "openvibe/economy"

The lowest cost per successful answer.

Fastestmodel: "openvibe/fast"

The quickest measured offer.

Privatemodel: "openvibe/private"

Only capacity OpenVibe runs or that sits on a local machine.

Balancedmodel: "openvibe/auto"

Cost, health and latency weighed together. The default.

These four aliases work today. A saved preset (model: "@preset/private") is the router’s floor, ceiling, quantization, region and data policy written down once. The private preset is the strict one: only offers that keep nothing and train on nothing, or OpenVibe’s own deployments, may answer, and the router refuses rather than fall back.

Try it

Chat with DeepSeek V4.1 Flash How it ranks

Use it from code

curl https://api.openvibe.exchange/v1/chat/completions \
  -H "Authorization: Bearer $OPENVIBE_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"deepseek/deepseek-flash","messages":[{"role":"user","content":"Hello"}]}'
import OpenAI from 'openai';
const ai = new OpenAI({ baseURL: 'https://api.openvibe.exchange/v1', apiKey: process.env.OPENVIBE_KEY });
const r = await ai.chat.completions.create({ model: 'deepseek/deepseek-flash', messages: [{ role: 'user', content: 'Hello' }] });

The endpoint, keys and usage: the API.