DeepSeek

DeepSeek V4.1 Flash

Launched 2026-09-10 · Knowledge up to 2025-05 · Listed on OpenRouter 2026-09-10

Prices updated . Benchmarks updated .

DeepSeek 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...

You send / 1M

$0.140

tokens in

It replies / 1M

$0.420

tokens out

Context

1M

tokens it can read at once

Typical request

$0.0004

1,000 words in, 400 out

Benchmarks

BenchmarkScoreKindDateSource
Epoch Capabilities Index156.1Independent2026-09-09Epoch AI
GPQA DiamondMax effort89.8%Independent2026-10-09Epoch AI
OTIS Mock AIMEMax effort98.3%Independent2026-10-08Epoch AI
SimpleQA VerifiedMax effort42.9%Independent2026-10-13Epoch AI
ARC-AGI-2Max effort72.9%Independent2026-09-09ARC Prize, republished by Epoch AI
GDP.pdfMax effort19.8%Independent2026-09-09Surge AI GDP.pdf, republished by Epoch AI
CritPtMax effort14.3%Independent2026-09-09CritPt, republished by Epoch AI

Rankings on this site use the Epoch Capabilities Index, an independent composite. Lab-reported rows are the lab's own figures and are kept separate.

Data privacy

No training or retention: Yes. At least one provider serves it with zero data retention.

ProviderYou send / 1MIt replies / 1MUptimePrivacy policy
DeepInfra · fp8$0.112$0.336100%Read
River$0.096$0.384100%Read
DeepInfra · fp8$0.140$0.420100%Read
Novita · fp8$0.120$0.480100%Read
Relace · fp4$0.120$0.500100%Read
Wafer · mxfp4$0.051$0.567100%Read
InferenceNet · fp8$0.103$0.600100%Read
InferenceNet · fp8$0.103$0.600100%Read
DeepInfra · fp8$0.210$0.630100%Read
Reka$0.120$0.64099%Read
CoreWeave · fp8$0.130$0.650100%Read
Wafer · mxfp4$0.190$0.650100%Read
Cohere · fp8$0.195$0.660100%Read
Mancer 2 · fp8$0.212$0.750100%Read
Fireworks$0.210$0.840100%Read
Cloudflare$0.100$0.900100%Read
DigitalOcean$0.225$0.900100%Read
Makora · fp8$0.260$1.17100%n/a
Google$0.300$1.2096%Read
Crusoe · fp8$0.290$1.20100%Read
OpenInference · fp4$0.090$1.20100%Read
BaseTen$0.300$1.20100%Read
Together$0.300$1.20100%Read
SiliconFlow · fp8$0.300$1.20100%Read
Modal$0.300$1.20100%Read
BaseTen$0.300$1.20100%Read
Parasail · fp8$0.300$1.20100%Read
Fireworks$0.300$1.20100%Read
Venice · fp8$0.300$1.20100%Read
Parasail · fp8$0.450$1.80100%Read
Fireworks$0.450$1.80100%Read
Venice · fp8$3.00$12.00100%Read

Based on OpenRouter and Vercel AI Gateway data. Confirm terms with the provider before sending sensitive data.

What it can do

  • Reads: text, image
  • Writes: text
  • Weights: Open weights
  • Reasoning: Yes
  • Tool calling: Yes
  • Longest reply: 384K tokens
  • Repeated text is billed at $0.0042 per 1M (prompt caching)

Compare with

Appears in

Estimate your monthly cost with DeepSeek V4.1 Flash

Sources and licenses

  • OpenRouter: Prices, context windows, modalities and the model list come from the OpenRouter public API. Data: OpenRouter.
  • Vercel AI Gateway: The no-training flag comes from the Vercel AI Gateway model list.
  • models.dev: Lab launch dates, open-weights flags and knowledge cutoffs come from models.dev. Copyright (c) 2025 models.dev, MIT License.
  • Epoch AI: Independent benchmark scores and the Epoch Capabilities Index (ECI) come from Epoch AI's Benchmarking Hub, used under CC BY 4.0. Epoch AI, Capabilities & benchmarking, epoch.ai/benchmarks. Scores Epoch marks as external keep their original license, and the source is named on each score.
  • LMArena: Arena Elo (text, webdev and vision) comes from the LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset), CC BY 4.0, leaderboard date 2026-10-28.
  • Lab-reported scores, when an editor has added them, are the lab's own numbers. Each one links to the source and is labeled lab-reported. They are not used to rank the lists.
  • What a score measures: the benchmark guide.