All promotions
GuidesMCP

DeepSeek V4.1 Flash

DeepSeekReleased Sep 10, 2026Open weights · MITdeepseek-v4.1-flash

Compare with other models
Intelligence (Epoch)
156 · 24th of 109
Reasoning
yes, with effort levels
Knowledge cutoff
March 2026
Context
1M tokens
Max output
384K tokens
Input
text, image
API features
tools, structured outputs, JSON mode, seed, logprobs

Read Nov 3, 11:57 UTC from the hosts' published offers and OpenRouter, spec sheet from DeepSeek · cutoff: aiknowledgecutoff.com

Lowest input, per M$0.035at Morph
Cheapest cache, per M$0.002at CoreWeave
Fastest252 tok/sat CoreWeave, on OpenRouter
Providers25serve this version
Sorted by cache price
ProviderInputOutputCache ↑tok/sFirst tokenUptime 24hChecked
CoreWeavefp8 · 1M ctx$0.13$0.65$0.0022520.51 s100%Nov 3
StreamLakefp8 · 1000K ctx · 375K out−51%$0.15$0.59$0.0029731.25 s100%Nov 3
Cohereprecision not read: host terms · 1M ctx · 375K out$0.30$1.20$0.0031950.40 s100%Nov 3
GMICloudfp8 · 1.05M ctx · 375K out−40%$0.18$0.72$0.00361652.36 s96.8%Nov 3
DeepInfrafp8 · 1M ctx · 384K out−30%$0.14$0.42$0.0042920.61 s100%Nov 3
Phalaprecision not read: host terms · 1M ctx · 384K out−30%$0.21$0.84$0.0042940.98 s99.8%Nov 3
Morphfp8 · 1M ctxreprices often$0.035$0.32$0.006630.65 s99.8%Nov 3
Novita AIfp8 · 1M ctx · 384K out$0.30$1.20$0.006———Nov 3
Fireworks AI1M ctx · 375K out$0.30$1.20$0.0061110.65 s97.7%Nov 3
Together AIfp8 · 1M ctx$0.30$1.20$0.0062230.29 s100%Nov 3
SiliconFlowfp8 · 1M ctx · 384K out$0.30$1.20$0.0061011.35 s100%Nov 3
Parasailfp8 · 1M ctx · 375K out$0.30$1.20$0.0061750.53 s100%Nov 3
DigitalOcean1M ctx$0.30$1.20$0.0061590.71 s100%Nov 3
Baidufp8 · 1M ctx · 384K outreprices often$0.30$1.20$0.006810.89 s100%Nov 3
Rekaprecision not read: host terms · 1M ctx · 384K out−30%$0.10$0.84$0.007293.59 s98.1%Nov 3
InferenceNetfp8 · 1.04M ctx · 384K out$0.30$1.00$0.007940.31 s100%Nov 3
Crusoefp8 · 1M ctx$0.29$1.20$0.0071500.77 s99.2%Nov 3
BaseTenprecision not stated · 1M ctx · 256K out$0.30$1.20$0.0072160.52 s99.9%Nov 3
Venicefp8 · 1M ctx · 128K out$0.30$1.20$0.00751710.45 s99.9%Nov 3
Relacefp4 (fp8) · 1M ctxreprices often$0.07$0.50$0.0081430.49 s99.2%Nov 3
Nebiusfp8 · 1M ctx$0.30$1.20$0.01590.70 s100%Nov 3
AtlasCloudfp8 · 1M ctx · 384K out$0.14$0.54$0.0135680.75 s100%Nov 3
Google1M ctx$0.30$1.20$0.031070.64 s97.0%Nov 3
Alibabaprecision not read: host terms · 1M ctx · 384K outpeak hours$0.33$1.32$0.0331170.71 s100%Nov 3

10 of 25 providers · prices per million tokens, read Nov 3, 11:57 UTC · speed: OpenRouter, median of the last 30 min, read Nov 3, 11:57 UTC · uptime: OpenRouter, last 24 h · −N%: a discount off the host's list price · under each host: weights precision, context and max output as served, the maker's value in brackets when the host's differs

ProviderWeightsContextMax outputToolsStructured outputsReasoning
Morphfp81Mnot stated by the hostyesyesyes
DeepInfrafp81M384Kyesyesyes
Relacefp41Mnot stated by the hostyesyesyes
AtlasCloudfp81M384Kyesyesyes
StreamLakefp81000K375Kyesnoyes
CoreWeavefp81Mnot checked yetyesyesyes
GMICloudfp81.05M375Kyesnoyes
Rekanot read: host terms1M384Kyesyesyes
Phalanot read: host terms1M384Kyesnoyes
InferenceNetfp81.04M384Kyesyesyes
Crusoefp81Mnot checked yetyesyesyes
Novita AIfp81M384Kyesnoyes
Fireworks AInot checked yet1M375Kyesyesyes
Together AIfp81Mnot stated by the hostyesyesyes
SiliconFlowfp81M384Kyesnoyes
Modalnot stated by the host1Mnot stated by the hostyesyesyes
BaseTennot stated by the host1M256Kyesyesyes
Parasailfp81M375Kyesyesyes
Venicefp81M128Kyesyesyes
DigitalOceannot checked yet1Mnot checked yetyesyesyes
Baidufp81M384Kyesyesyes
Nebiusfp81Mnot stated by the hostyesyesyes
Googlenot checked yet1Mnot checked yetyesyesyes
Coherenot read: host terms1M375Kyesyesyes
Alibabanot read: host terms1M384Kyesnoyes

Weights, context, max output: from the host where its terms allow, else as declared to OpenRouter (hover for source and date) · features as declared to OpenRouter, not verified by us

DeepSeek V4.1 Flash spec sheet, field by field, with sources

DeepSeek V4.1 Flash spec sheetmaker-declared

What DeepSeek declares, from its own weights and docs. What each host serves is in the offers: both can be true.

Context window
1,048,576 tokens
Parameters
763,205,315,794≈ 763.2B
Active experts
6 of 384 active per token
Weight precision
fp8
License
MIT — open weights, permissivecurated · read 2026-09-10Scope: the license of the published weights only — not the maker's acceptable-use policy, nor the terms of the host you call.
Knowledge cutoff
not published by the maker≈ March 2026 third-party estimate — not the maker's figuremaker's docs · read 2026-09-10
Modalities
text · image
Max output (official)
384,000 tokensmaker's docs · read 2026-09-10
Release date
September 10, 2026maker's docs · read 2026-09-10
Maker lifecycle status
General availabilitymaker's docs · read 2026-09-10

Also declared in the weights

Architecture class
DeepseekV41ForCausalLM
Layers
40
Vocabulary
129,280 tokens
Weights first published
September 10, 2026
Weights last updated
October 1, 2026

deepseek-ai/DeepSeek-V4.1-Flash · config.json @ 2cba9e4 · read 2026-10-08 · 9 of 10 fields published by the maker

DateProviderInputOutputChange
Nov 3Morph$0.05 → $0.035$0.19 → $0.32+45%
Nov 2Baidu$0.0447 → $0.30$0.18 → $1.20+571%
Nov 1Baidu$0.30 → $0.0447$1.20 → $0.18−85%
Nov 1Morph$0.022 → $0.05$0.19 → $0.19+13%
Oct 31Morph$0.044 → $0.022$0.38 → $0.19−50%
Oct 30Relace$0.20 → $0.07$0.60 → $0.50−29%
Oct 29Morph$0.09 → $0.044$0.78 → $0.38−51%
Oct 28Morph$0.069 → $0.09$0.60 → $0.78+30%
Oct 27CoreWeave$0.15 → $0.13$0.65 → $0.65−3%
Oct 27Morph$0.10 → $0.069$0.60 → $0.60−4%
Oct 26Baidu$0.0537 → $0.30$0.21 → $1.20+459%
Oct 25Morph$0.45 → $0.10$0.60 → $0.60−34%
Oct 24Morph$0.024 → $0.45$0.60 → $0.60+69%
Oct 23Baidu$0.075 → $0.0537$0.30 → $0.21−28%
Oct 23Morph$0.25 → $0.024$0.60 → $0.60−27%
Oct 22Baidu$0.30 → $0.075$1.20 → $0.30−75%
Oct 22Fireworks AI$0.21 → $0.30$0.84 → $1.20+43%
Oct 22Reka$0.14 → $0.10$0.56 → $0.84+34%
Oct 20Alibaba$0.30 → $0.33$1.20 → $1.32+10%
Oct 20Fireworks AI$0.30 → $0.21$1.20 → $0.84−30%
Oct 20StreamLake$0.15 → $0.15$0.60 → $0.59−2%
Oct 17AtlasCloud$0.30 → $0.14$1.20 → $0.54−55%
Oct 16AtlasCloud$0.19 → $0.30$0.77 → $1.20+56%
Oct 16StreamLake$0.15 → $0.15$0.59 → $0.60+2%
Oct 15Baidu$0.17 → $0.30$0.68 → $1.20+75%
Oct 15CoreWeave$0.20 → $0.15$0.65 → $0.65−6%
Oct 14AtlasCloud$0.30 → $0.19$1.20 → $0.77−36%
Oct 14Baidu$0.30 → $0.17$1.20 → $0.68−43%
Oct 14Cloudflare$0.30 → $0.10$1.20 → $0.90−33%
Oct 12Baidu$0.15 → $0.30$0.60 → $1.20+100%
Oct 11Baidu$0.30 → $0.15$1.20 → $0.60−50%
Oct 11StreamLake$0.15 → $0.15$0.60 → $0.59−2%
Oct 9AtlasCloud$0.14 → $0.30$0.56 → $1.20+113%
Oct 9StreamLake$0.14 → $0.15$0.58 → $0.60+4%
Oct 8DekaLLM$0.12 → $0.12$2.40 → $1.20−48%
Oct 8StreamLake$0.14 → $0.14$0.56 → $0.58+2%
Oct 5DekaLLM$0.12 → $0.12$0.40 → $2.40+385%
Oct 3Baidu$0.30 → $0.30$1.20 → $1.200%
Oct 3Venice$0.38 → $0.30$1.50 → $1.20−20%
Oct 2Fireworks AI$0.22 → $0.30$0.66 → $1.20+70%
Oct 2Io Net$0.16 → $0.099$0.51 → $0.37−30%
Oct 2Reka$0.29 → $0.14$1.16 → $0.56−51%
Oct 2StreamLake$0.14 → $0.14$0.58 → $0.56−2%
Oct 1AtlasCloud$0.11 → $0.14$0.46 → $0.56+24%
Oct 1Baidu$0.30 → $0.30$1.20 → $1.200%
Oct 1Io Net$0.13 → $0.16$0.41 → $0.51+24%
Oct 1StreamLake$0.15 → $0.14$0.59 → $0.58−2%
Sep 30StreamLake$0.18 → $0.15$0.71 → $0.59−18%
Sep 29AtlasCloud$0.30 → $0.11$1.20 → $0.46−62%
Sep 28DekaLLM$0.15 → $0.12$0.60 → $0.40−31%
Sep 28StreamLake$0.17 → $0.18$0.66 → $0.71+8%
Sep 27GMICloud$0.23 → $0.18$0.90 → $0.72−20%
Sep 26DekaLLM$0.04 → $0.15$0.49 → $0.60+42%
Sep 25StreamLake$0.20 → $0.17$0.79 → $0.66−17%
Sep 24GMICloud$0.28 → $0.23$1.14 → $0.90−21%
Sep 24StreamLake$0.15 → $0.20$0.60 → $0.79+32%
Sep 22Alibaba$0.15 → $0.30$0.60 → $1.20+100%
Sep 18Alibaba$0.30 → $0.15$1.20 → $0.60−50%
Sep 16GMICloud$0.30 → $0.28$1.20 → $1.14−5%

Every published change of input or output price, newest first · change on input + output · full series: pricing-history.csv.

ProviderChatbot 1,000 in / 500 outRAG / long context 8,000 in / 500 outBatch summarize 4,000 in / 1,000 out
Morph$0.19$0.44$0.46
DeepInfra$0.35$1.33$0.98
Relace$0.32$0.81$0.78
AtlasCloud$0.40$1.35$1.08
StreamLake$0.44$1.47$1.18
CoreWeave$0.46$1.36$1.17
GMICloud$0.54$1.80$1.44
Reka$0.53$1.26$1.26
Phala$0.63$2.10$1.68
InferenceNet$0.80$2.90$2.20
Crusoe$0.89$2.92$2.36
Novita AI$0.90$3.00$2.40
Fireworks AI$0.90$3.00$2.40
Together AI$0.90$3.00$2.40
SiliconFlow$0.90$3.00$2.40
Modal$0.90$3.00$2.40
BaseTen$0.90$3.00$2.40
Parasail$0.90$3.00$2.40
Venice$0.90$3.00$2.40
DigitalOcean$0.90$3.00$2.40
Baidu$0.90$3.00$2.40
Nebius$0.90$3.00$2.40
Google$0.90$3.00$2.40
Cohere$0.90$3.00$2.40
Alibaba$0.99$3.30$2.64

What each host bills for 1,000 requests of a representative job

Your own workload

    Not seen at their source lately

    These hosts no longer appear where their price is read, so they are left out of every comparison on this page. The last price read is kept for the record.

    ProviderInputOutputCacheLast readStatus
    Cloudflare$0.10$0.90$0.03Oct 17not in Cloudflare's own catalogue since Oct 18
    NextBit$0.21$0.84$0.004Oct 6not seen where its price is read since Oct 7
    DekaLLM$0.12$1.20$0.005Oct 11not seen where its price is read since Oct 12
    No longer offered (1)
    • Io Net · last price $0.099 in / $0.37 out, read Oct 2 · price no longer verifiable

    Intelligence for the price

    On the best-deal line: no model scores higher for less than $0.35 per M.

    100120140160180$0.10$1$10$100DeepSeek V4.1 Flash · 156

    Epoch Capabilities Index against the lowest input + output price per M, log scale · prices Nov 3. Score: Epoch AI, CC BY 4.0, 156 with an interval of 154 to 158.

    Against its rivals

    Its previous version, the best of its price range and the top score

    Bars: Epoch Capabilities Index, drawn from 100 · price: lowest input + output per M.

    Speed by provider

    Output speed of DeepSeek V4.1 Flash at each host, fastest first (the 10 fastest of 24)

    Speed: OpenRouter, median of the last 30 min, read Nov 3, 11:57 UTC · a snapshot of real traffic before the daily read, not a daily average · bars from zero.

    Run it yourself

    open weights, MIT · fits 1 of the 14 workstations in our guideSelf-host guide ›
    Parameters
    552B
    16B active per token (MoE)
    Memory needed
    331 GB
    in INT4 · 662 GB in INT8
    Smallest cloud setup
    8× H100
    640 GB · $21.12/h, about $15,418 a month
    Worth it above
    44B tokens
    a month, against the cheapest API at $0.35 per M
    QuantizationMemorySmallest cloud GPUQuality
    INT4 (Q4)331 GB8× H100 · $21.12/hsmallest footprint, minor quality loss
    INT8 (Q8)662 GBmulti-node, over 640 GBnear-lossless
    FP16 (full)1325 GBmulti-node, over 640 GBreference quality

    At home: 331 GB in INT4 is beyond every machine we price, and the closest is the 512 GB Mac Studio, whose price Apple hasn't announced, so it means a multi-GPU server.

    Memory ≈ parameters × bytes × 1.2 · cloud: on-demand, 24/7 · builds: GGUF, AWQ, MLX, GPTQ, Ollama · break-even calculator

    Other DeepSeek models · all its models, hosts and history ›