ModelPriceWatch.com
Last scan 2026-10-25 Models tracked 301 Providers 40 Cheapest paid pplx-decider-flash-v1.2 $0.010/Mtok in Every price links to its source

DeepSeek V4.1 Flash

by DeepSeek · 552B (A8B/A16B) parameters

Retired budget open weights cheap tier
Email me when DeepSeek V4.1 Flash gets cheaper
Retired on . No longer served on the endpoint we price — the figures below are kept for historical reference only. Replaced by DeepSeek V4.1 Flash 1017. Full retirement status & replacements →
Availability: Retired 2026-10-17: superseded by DeepSeek V4.1 Flash 1017 (deepseek-v4-1-flash-1017). That day the MODEL VERSION column of DeepSeek's rate card changed from DeepSeek-V4.1-Flash to DeepSeek-V4.1-Flash-1017, and DeepSeek's release note (news261017) says the new build has stronger benchmark performance, faster task completion and shorter outputs than this one. The API name deepseek-flash did not change and neither did the six rates, so a call to deepseek-flash today reaches the 1017 build at the same per-token price archived here; this September build is no longer served under any name. The retired names deepseek-v4-flash and deepseek-v4-flash-vision-exp, which routed here from 2026-09-10, now route to the 1017 build too. The weights of this build stay downloadable from Hugging Face (deepseek-ai/DeepSeek-V4.1-Flash, MIT). DeepSeek had announced that deepseek-v4-pro would route here from 2026-09-14, but withdrew that on 2026-09-11 — V4 Pro keeps its own service and billing.

Today's price · per 1M tokens

Input

$0.300

per 1M tokens

Output

$1.20

per 1M tokens

Blended

$0.525

blended $/1M — 3:1 weighted input:output

Cached input

$0.006

2% of input — prompt caching

How this price is scoped: ARCHIVED first-party rate — what this build last cost. DeepSeek V4.1 Flash 1017, which replaced it behind the same deepseek-flash name on 2026-10-17, charges the identical schedule. DeepSeek billed this model at TWO rates depending on the hour. The tracked figure is the PEAK rate: $0.30 input (cache miss) / $0.006 input (cache hit) / $1.20 output per 1M tokens. Off-peak was exactly half: $0.15 / $0.003 / $0.60. Peak hours are 01:00-04:00 and 06:00-10:00 UTC, Monday through Friday; every other hour, and all of Saturday and Sunday, is off-peak.

Source: official DeepSeek pricing · read MODELPRICEWATCH.COM · 2026-10-25

Price receipt

We read DeepSeek's own pricing page on and found DeepSeek-V4.1-Flash listed with a price on the same column — that read is where the number above comes from. We keep the page text we read; its content hash is b34099bace.

Overview

Retired by DeepSeek on 2026-10-17, when DeepSeek-V4.1-Flash-1017 took over the deepseek-flash name. This is the September 2026 build of V4.1 Flash, DeepSeek's first model on its causal encoder-decoder architecture: a 552B-parameter MoE activating 8B parameters for input and 16B for output, with native visual understanding, a 1M-token context, 384K max output and open weights under the MIT licence. The deepseek-flash name is still accepted, but every request to it is now served by the 1017 build and billed at that model's price, so what you pay today is on the DeepSeek V4.1 Flash 1017 page. Last listed at $0.30/$1.20 per 1M tokens at peak, the rate its successor still charges.

Run it yourself

Deploy this open model on rented GPUs

Open weights mean you can self-host instead of paying per-token API prices. These platforms let you serve it on demand — often cheaper at scale.

RunPod →
Rent GPUs by the second and serve any open-weight model yourself.
Novita AI →
Serverless open-model inference billed per token — no GPU ops.
Vast.ai →
GPU marketplace — rent spot/on-demand GPUs cheaply and serve any open-weight model.

Partner links — we may earn a commission if you sign up, at no cost to you. We only list platforms we'd recommend regardless.

Capabilities

struck through = not supported
Input 2/4
Text ✓ Image ✓ Audio Video
Output
Text ✓
Features 6/9
Prompt caching ✓ Reasoning ✓ Coding Fast inference ✓ Long context ✓ Open weights ✓ Multimodal ✓ Web search Realtime

Benchmark performance

accuracy % · higher is better
Avg benchmark score56.8%
Perf per $/Mtok108.2
GPQA Diamond vendor
90.9%
Terminal-Bench 2.1
74.5%
Humanity's Last Exam
39%
Independent composite scores
  • AA Intelligence Index: 39.5 (Artificial Analysis composite across reasoning, knowledge and coding evals)
  • AA-Omniscience Index: -5.3 (−100–100)
  • GDPval-AA v2: 1600 ELO (Real-world work tasks, human baseline = 1000)
How it stacks up
  • Ranks #74 of 99 benchmarked models by average score
  • Ranks #53 of 194 comparably-measured models by percentile score, across 5 independent measurements
  • Ranks #21 of 194 comparably-tested models by normalized performance per dollar
  • Strongest at Terminal-Bench 2.1 — 74.5%, #25 of 46

214.4 tokens/sec output

Source: Artificial Analysis, Artificial Analysis (text-only, no tools, max effort), Vals AI (Terminal-Bench 2.1, Terminus 2, 3 trials) · updated · See full rankings →

Specifications

Provider
DeepSeek
Context window
1M tokens
Modality
text, image
Parameters
552B (A8B/A16B)
Open source
Yes — open weights available
Released
Status
Retired ended
Last updated
Tags
fastreasoningmultimodalcachingbudgetopen-weightsretired

Availability verified: — per DeepSeek's own deprecation notice

Try on DeepSeek Official pricing page