Skip to main content

    DeepSeek

    DeepSeek V4.1 Flash

    DeepSeek V4.1 Flash was released September 10, 2026 under an MIT license and is served through the deepseek-flash alias; the legacy deepseek-v4-flash slug temporarily routes to it. Its 552B-parameter backbone uses 8B active parameters for prefill and 16B for decode, with vision input and a 1M-token context. The 1017 checkpoint shipped October 17, 2026. Hosted access has peak-hour cache-miss pricing of $0.30 input and $1.20 output per million tokens, halved off-peak, with cache hits at $0.006.

    1M tokens
    Multimodal
    Pro
    New
    Provider
    DeepSeek
    Context
    1M
    Catalog · As of Oct 21, 2026Under 1d old
    yno subscription tier
    Pro
    Released
    Sep 10, 2026
    Speed
    Fast
    Reasoning
    Expert
    Modality
    Multimodal

    Benchmarks

    Quality

    AA Intelligence Index
    39.5points
    Artificial Analysis · Date unknownAge unknown

    Speed and latency

    Output speed
    206.17tokens/s
    Configuration: Prompt length: 1,000 · Parallel queries: 1
    Artificial Analysis · Date unknownAge unknown
    Time to first token
    0.73s
    Configuration: Prompt length: 1,000 · Parallel queries: 1
    Artificial Analysis · Date unknownAge unknown
    Time to first answer token
    10.44s
    Configuration: Prompt length: 1,000 · Parallel queries: 1
    Artificial Analysis · Date unknownAge unknown

    API pricing

    API input cost
    $0.30/ 1M tokens
    Artificial Analysis · Date unknownAge unknown
    API output cost
    $1.20/ 1M tokens
    Artificial Analysis · Date unknownAge unknown

    Benchmark variant: DeepSeek V4.1 Flash (Max)

    AA data retrieved Oct 21, 2026 · Artificial Analysis

    Indexes use points; evaluations use accuracy percentages. API measurements do not measure yno application performance. Methodology

    Capabilities

    • Open Source
    • Reasoning
    • Vision
    • Long Context

    Use cases

    • Multimodal applications
    • Document processing
    • Self-hosted AI
    • High-volume APIs

    Strengths

    • MIT-licensed open weights
    • 552B backbone with separate prefill/decode activation
    • 1M-token context
    • Vision-capable successor to V4 Flash

    Best for

    Teams moving V4 Flash workloads to a vision-capable model with hosted or self-hosted access

    Use DeepSeek V4.1 Flash in yno.ai

    No credit card required

    Related models

    DeepSeek V4 Pro

    DeepSeek

    DeepSeek V4 Flash

    DeepSeek

    DeepSeek V3.2

    DeepSeek

    Frequently asked

    What is DeepSeek V4.1 Flash?
    DeepSeek V4.1 Flash was released September 10, 2026 under an MIT license and is served through the deepseek-flash alias; the legacy deepseek-v4-flash slug temporarily routes to it. Its 552B-parameter backbone uses 8B active parameters for prefill and 16B for decode, with vision input and a 1M-token context. The 1017 checkpoint shipped October 17, 2026. Hosted access has peak-hour cache-miss pricing of $0.30 input and $1.20 output per million tokens, halved off-peak, with cache hits at $0.006.
    How much does DeepSeek V4.1 Flash cost in yno.ai?
    DeepSeek V4.1 Flash is available on the Pro tier of yno.ai.
    What can DeepSeek V4.1 Flash do?
    DeepSeek V4.1 Flash is best for Teams moving V4 Flash workloads to a vision-capable model with hosted or self-hosted access. Its main capabilities include Open Source, Reasoning, Vision, Long Context.
    How does DeepSeek V4.1 Flash compare to other models?
    DeepSeek V4.1 Flash excels at MIT-licensed open weights and is recommended for Multimodal applications, Document processing, Self-hosted AI, High-volume APIs. See related models below.
    How do I use DeepSeek V4.1 Flash in yno.ai?
    Sign up for yno.ai, select DeepSeek V4.1 Flash from the model picker, and start chatting.