Skip to Sidebar Skip to Content
Company tracker 7 of 10 · 10 companies, one lens

DeepSeek

The maker of the DeepSeek models, tracked through the same six lenses as every company on AI Adoption.
Its best model, DeepSeek V4.1 Flash, scores 155.9 on the Epoch Capabilities Index, #7 of 10, 12.9 behind Anthropic. Annualized revenue reached $1B by 24 Sep 2026, and its Jun 2026 round valued it at $50.0B. In March 2026, 0.6% of US adults said they had used DeepSeek in the past week. It released 4 notable models in 2025, #6 of 10 by count.
Follow DeepSeekCompare with the others
35 ledger entries4 charts name DeepSeekLatest entry Sep 2026Checked 20 Oct 2026

DeepSeek through six lenses

How the lenses work →
The same six lenses, in the same order, on every company page. Each shows one comparable number where a source measures several companies the same way, its rank, and when it last changed.
01Capability#7 of 10
How good are its best models?
155.9
Epoch Capabilities Index score of its best model, DeepSeek V4.1 Flash (Sep 2026)
Anthropic168.8OpenAI168.8Google166.4Meta161.8DeepSeek155.9
Epoch combines many benchmarks into one scale. 90% interval for this model: 153.7 to 158.6.
Text Arena, Mar 2026: 1,424, #6 of 8. Arena chart →
Independent measureChanged Sep 2026Epoch AI data →
02Output#6 of 10
How much does it ship?
4
Notable AI models released in 2025
OpenAI21Google15Alibaba11Anthropic7DeepSeek4
7 notable models from 2014 to 2025 and 3 in 2026 so far. Model record →
Independent measureChanged 2025View chart →
03Money#4 of 9
How much does it earn?
$1B
Annualized revenue, 24 Sep 2026
Anthropic$65BOpenAI$50BZ.ai (Zhipu)$1.6BDeepSeek$1B
Annualized run rate; Epoch rates the report “Likely”. Dates differ by company; each is its latest report. Ranked among the 9 companies in Epoch AI's revenue data.
Valuation $50.0B after its Jun 2026 round; $7.5B raised in equity in all; about 300 staff (Jun 2026).
Reported figureChanged Sep 2026Epoch AI data →
04Compute#8 of 10
How much compute and energy goes into its AI?
9.7×10²⁴
FLOP: its largest training run with a published estimate, DeepSeek-V4-Pro (Apr 2026)
OpenAI9.4×10²⁶xAI5×10²⁶Meta5.2×10²⁵Google5×10²⁵DeepSeek9.7×10²⁴
Largest estimated training cost: $7M (DeepSeek-R1 (May 2025)), Epoch AI. Model record →
Ranked among the 10 tracked companies. Only runs with a published estimate count, so a company that discloses less can rank lower.
Independent estimateChanged Apr 2026View chart →
05Use#7 of 7
Who uses it, and for what?
0.6%
of US adults said they used DeepSeek in the past week, March 2026
OpenAI31.0%Google20.9%Microsoft10.5%Meta7.5%DeepSeek0.6%
Epoch AI with Ipsos, a probability-based sample of 2,021 US adults; 49.2% used no AI service.
Independent measureChanged Mar 2026Epoch AI data →
06Safety & policy#4 of 9
How reliable is it, and how is it governed?
82%
hallucination rate of its most reliable model on AA-Omniscience, DeepSeek V3.2
xAI22%Anthropic26%Google50%DeepSeek82%
Lower is better. Ranked among the tracked companies; the test covers 26 models. One reliability test: safety frameworks, evaluations and regulatory actions are logged as entries.
Independent measureChanged 2026View chart →

The ledger

What gets an entry →
An entry goes in only if it changes a number or a position on one of the six lenses. If it moves nothing, it is not news here. Newest first.

  1. September 2026

    1. 24 Sep 2026MoneyNewAnnualized revenue reached $1B, from $450M on 14 Jul 2026Annualized run rate; Epoch rates the report “Likely”.Reported figureEpoch AI, AI CompaniesEpoch AI data →+$550M
    2. 9 Sep 2026CapabilityNewDeepSeek V4.1 Flash scored 155.9 on the Epoch Capabilities Index, its best model#7 of the 10 tracked companies' best models; Anthropic's best scored 168.8. Epoch's 90% interval: 153.7 to 158.6.Independent measureEpoch AI, Benchmarking HubEpoch AI data →#7
    3. 9 Sep 2026OutputNewDeepSeek V4.1 Flash released, trained with an estimated 4.3×10²⁴ FLOPLess training compute than DeepSeek-V4-Pro (9.7×10²⁴).Independent estimateEpoch AIView chart →4.3×10²⁴

  2. June 2026

    1. 16 Jun 2026MoneyNewClosed a funding round at a $50.0B post-money valuation, raising $7.5B in equityReported figureEpoch AI, AI CompaniesEpoch AI data →$50.0B

  3. April 2026

    1. 24 Apr 2026OutputNewDeepSeek-V4-Pro released, trained with an estimated 9.7×10²⁴ FLOPAbout 3× its previous model in this data, DeepSeek-R1 (Jan 2025).Independent estimateEpoch AIView chart →9.7×10²⁴

  4. March 2026

    1. Mar 2026UseNew0.6% of US adults said they had used DeepSeek in the past week#7 of the 7 services asked about that belong to tracked companies. 49.2% had used no AI service. Probability-based sample of 2,021 US adults.Independent measureEpoch AI and IpsosEpoch AI data →#7

  5. 2026 tests

    1. 2026 testSafety & policyNewIts most reliable model on AA-Omniscience, DeepSeek V3.2, had a 82% hallucination rate#4 of 9 tracked companies; xAI's best was 22%. Rates across the 26 models tested ran from 22% to 94%.Independent measureArtificial AnalysisView chart →#4 of 9

  6. March 2026

    1. Mar 2026CapabilityNewStill #6 of 8 with 1,424; the top four labs sat within 22 Arena pointsAnthropic 1,503, xAI 1,495, Google 1,494, OpenAI 1,481. In May 2023 the leader was 205 points clear.Independent measureArenaView chart →#6

  7. 2025, full year

    1. 2025OutputNewReleased 4 notable AI models in 2025, #6 by countOpenAI 21, Google 15, Alibaba 11. Epoch AI counts a model as notable if it was state of the art, highly cited or historically significant.Independent measureEpoch AIView chart →#6
    2. 2025ComputeNewDeepSeek V3.2 Exp used 23.2 Wh per medium-length prompt, the most of the 15 most energy-intensive 2025 modelsThe least of the 15, GPT-5 mini (medium), used 5.1 Wh.Independent estimateJegham et al.View chart →23.2 Wh

  8. November 2025

    1. Nov 2025CapabilityNewFell to #6 of 8 on the Text Arena, 68 points behind GoogleGoogle led at 1,492. First to fourth spanned 31 points.Independent measureArenaView chart →#5 → #6

  9. October 2025

    1. Oct 2025CapabilityNewRose to #5 of 8 on the Text Arena, 31 points behind GoogleGoogle led at 1,452. First to fourth spanned 19 points.Independent measureArenaView chart →#6 → #5

  10. September 2025

    1. Sep 2025CapabilityNewFell to #6 of 8 on the Text Arena, 35 points behind GoogleGoogle led at 1,453. First to fourth spanned 17 points.Independent measureArenaView chart →#4 → #6

  11. August 2025

    1. Aug 2025CapabilityNewRose to #4 of 8 on the Text Arena, 30 points behind GoogleGoogle led at 1,461. First to fourth spanned 30 points.Independent measureArenaView chart →#6 → #4

  12. July 2025

    1. Jul 2025CapabilityNewFell to #6 of 8 on the Text Arena, 65 points behind OpenAIOpenAI led at 1,482. First to fourth spanned 50 points.Independent measureArenaView chart →#3 → #6

  13. 2025 studies

    1. 2025 studyCapabilityNewDeepSeek R1 solved 157 PlanBench tasks, against 205 for GPT-5 and 204 for the classical planner LAMAGemini solved 155. LAMA led or tied in six of eight domains.Independent measureCorrêa et al.View chart →
    2. 2025 studyCapabilityNewDisguising PlanBench tasks cut DeepSeek R1's solved total from 157 to 93GPT-5's fell from 205 to 152; the classical planner LAMA still solved 204.Independent measureCorrêa et al.View chart →−64

  14. June 2025

    1. Jun 2025CapabilityNewRose to #3 of 8 on the Text Arena, 49 points behind GoogleGoogle led at 1,472. First to fourth spanned 52 points.Independent measureArenaView chart →#5 → #3

  15. May 2025

    1. May 2025CapabilityNewFell to #5 of 8 on the Text Arena, 91 points behind GoogleGoogle led at 1,492. First to fourth spanned 80 points.Independent measureArenaView chart →#3 → #5

  16. April 2025

    1. Apr 2025CapabilityNewRose to #3 of 8 on the Text Arena, 57 points behind OpenAIOpenAI led at 1,461. First to fourth spanned 70 points.Independent measureArenaView chart →#4 → #3

  17. March 2025

    1. Mar 2025CapabilityNewFell to #4 of 8 on the Text Arena, 46 points behind OpenAIOpenAI led at 1,441. First to fourth spanned 46 points.Independent measureArenaView chart →#2 → #4

  18. January 2025

    1. 20 Jan 2025OutputNewDeepSeek-R1 released, trained with an estimated 3.5×10²⁴ FLOPIndependent estimateEpoch AIView chart →3.5×10²⁴
    2. Jan 2025CapabilityNewRose to #2 of 8 on the Text Arena, 8 points behind OpenAIOpenAI led at 1,405. First to fourth spanned 56 points.Independent measureArenaView chart →#3 → #2

  19. 2024, full year

    1. 2024ComputeNewTraining DeepSeek v3 emitted an estimated 597 tonnes of CO₂, a fifteenth of Llama 3.1 405B's 8,930Despite having more parameters: 671B.Independent estimateAI IndexView chart →

  20. December 2024

    1. 24 Dec 2024OutputNewDeepSeek-V3 released, trained with an estimated 3.3×10²⁴ FLOPAbout 2× its previous model in this data, DeepSeek-V2.5 (Sep 2024).Independent estimateEpoch AIView chart →3.3×10²⁴
    2. Dec 2024CapabilityNewRose to #3 of 8 on the Text Arena, 47 points behind OpenAIOpenAI led at 1,407. First to fourth spanned 56 points.Independent measureArenaView chart →#8 → #3

  21. September 2024

    1. 6 Sep 2024OutputNewDeepSeek-V2.5 released, trained with an estimated 1.8×10²⁴ FLOPIndependent estimateEpoch AIView chart →1.8×10²⁴

  22. August 2024

    1. Aug 2024CapabilityNewFell to #8 of 8 on the Text Arena, 82 points behind OpenAIOpenAI led at 1,341. First to fourth spanned 10 points.Independent measureArenaView chart →#7 → #8

  23. July 2024

    1. Jul 2024CapabilityNewFell to #7 of 7 on the Text Arena, 82 points behind OpenAIOpenAI led at 1,341. First to fourth spanned 14 points.Independent measureArenaView chart →#6 → #7

  24. June 2024

    1. 17 Jun 2024OutputNewDeepSeek-Coder-V2 236B released, trained with an estimated 1.3×10²⁴ FLOPIndependent estimateEpoch AIView chart →1.3×10²⁴
    2. Jun 2024CapabilityNewRose to #6 of 7 on the Text Arena, 83 points behind OpenAIOpenAI led at 1,341. First to fourth spanned 71 points.Independent measureArenaView chart →#7 → #6

  25. April 2024

    1. Apr 2024CapabilityNewFell to #7 of 7 on the Text Arena, 153 points behind OpenAIOpenAI led at 1,338. First to fourth spanned 91 points.Independent measureArenaView chart →#6 → #7

  26. March 2024

    1. Mar 2024CapabilityNewFell to #6 of 7 on the Text Arena, 148 points behind AnthropicAnthropic led at 1,333. First to fourth spanned 95 points.Independent measureArenaView chart →#5 → #6

  27. February 2024

    1. Feb 2024CapabilityNewFell to #5 of 6 on the Text Arena, 143 points behind OpenAIOpenAI led at 1,329. First to fourth spanned 113 points.Independent measureArenaView chart →#4 → #5

  28. January 2024

    1. Jan 2024CapabilityNewPlaced #4 of 6 on the Text Arena with 1,189 points, its first month in this seriesOpenAI led at 1,334. First to fourth spanned 145 points.Independent measureArenaView chart →#4

DeepSeek's model record

Download the CSV →
Every notable model Epoch AI has logged for DeepSeek: 10 in all, 7 from 2014 to 2025 and 3 in 2026 so far. Training compute, cost and hardware are Epoch's estimates, shown only where it has one.
Notable models, 2014–20257#8 of 10 companies · Epoch AI's count, as in chart 1.1.7
Frontier models0#8 of 10 companies · marked frontier in Epoch AI's database
Open weights100%10 of 10 notable models since 2023
Largest estimated training cost$7M#8 of 10 companies · DeepSeek-R1 (May 2025) (May 2025), 2023 dollars
Largest training cluster2,048#7 of 8 companies · NVIDIA H800 SXM5 · DeepSeek-V3 (Mar 2025)
Notable models per year* 2026 to 19 Oct 2026
2014: 020142015: 02016: 020162017: 02018: 020182019: 02020: 020202021: 02022: 020222023: 02024: 3320242025: 442026: 332026*
Notable models since 2023, newest first10 models
ReleasedModelAccessTraining computeEst. costTraining hardware
9 Sep 2026DeepSeek V4.1 FlashOpen weights4.3×10²⁴——
24 Apr 2026DeepSeek-V4-ProOpen weights9.7×10²⁴——
24 Apr 2026DeepSeek-V4-FlashOpen weights2.5×10²⁴——
27 Nov 2025DeepSeekMath-V2Open weights———
28 May 2025DeepSeek-R1 (May 2025)Open weights4×10²⁴$7M—
24 Mar 2025DeepSeek-V3 (Mar 2025)Open weights3.3×10²⁴$5M2,048 NVIDIA H800 SXM5
20 Jan 2025DeepSeek-R1Open weights3.5×10²⁴$7M—
24 Dec 2024DeepSeek-V3Open weights3.3×10²⁴$5M2,048 NVIDIA H800 SXM5
6 Sep 2024DeepSeek-V2.5Open weights1.8×10²⁴——
17 Jun 2024DeepSeek-Coder-V2 236BOpen weights1.3×10²⁴——
Data: Epoch AI, Data on AI Models, snapshot 19 Oct 2026, CC BY 4.0. Organisations are grouped into companies as in chart 1.1.7 (for example, Google DeepMind and Google Brain count as Google). DeepSeek's models as CSV.

DeepSeek beside the others

All 10 companies →
The same rows for every company. A dash means no source measures it for that company yet: gaps are shown, never filled with a guess. The small number is the rank among the companies in the full table that the source covers.
LensDeepSeekOpenAIAnthropicGooglexAIMetaMicrosoftAlibabaMistral AINvidia
CapabilityEpoch Capabilities Index, its best model · data155.9#7168.8#1168.8#1166.4#3156.6#6161.8#4130.5#10157.4#5150.0#8146.0#9
OutputNotable AI models released in 2025 · chart4#621#17#415#25#54#60#1011#31#93#8
MoneyAnnualized revenue, latest report · data$1B#3Sep 2026$50B#2Sep 2026$65B#1Jul 2026—$428M#5Sep 2025———$1B#3Sep 2026—
ComputeLargest training run with a published estimate, FLOP · chart9.7×10²⁴#8DeepSeek-V4-Pro, Apr 20269.4×10²⁶#1GPT-6 Astra, Sep 20263.4×10²⁵#5Claude 3.7 Sonnet, Feb 20255×10²⁵#4Gemini 1.0 Ultra, Dec 20235×10²⁶#2Grok 4, Jul 20255.2×10²⁵#3Llama 4 Behemoth (preview), Apr 20258.6×10²³#9Megatron-Turing NLG 530B, Oct 2021 (with Nvidia)1.5×10²⁵#7Qwen3-Max, Sep 20257.7×10²³#10Mixtral 8x7B, Dec 20231.8×10²⁵#6Nemotron-4 340B, Jun 2024
UseUS adults who used its assistant in the past week, Mar 2026 · data0.6%#731.0%#13.0%#620.9%#24.5%#57.5%#410.5%#3———
Safety & policyLowest hallucination rate of its models (lower is better) · chart82%#489%#826%#250%#322%#187%#7—89%#884%#683%#5
More measures · Epoch AI data
Text Arenabest score, Mar 2026 · data1,424#61,481#41,503#11,494#31,495#21,335#8—1,449#51,416#7—
Valuationlatest closed round · data$50.0B#4$852.0B#2$965.0B#1—$230.0B#3———$24.0B#5—
Equity raisedin all · data$7.5B#4$182.8B#1$139.4B#2—$37.0B#3———$6.5B#5—
Stafflatest report · data300#64,500#23,000#46,000#1300#63,400#3——350#5—
AI chips ownedH100-equivalents · data———5.47M#10.55M#41.95M#33.32M#2———
Compute availableH100-equivalents, end of 2025 · data—1.74M#11.19M#31.58M#20.63M#51.00M#4————
Weekly use at workemployed US adults, Jul 2026 · data—33.7%#17.7%#524.2%#23.3%#69.8%#414.8%#3———
Model record · Epoch AI estimates
Notable models, 2014–2025Epoch AI's count, as in chart 1.1.77#861#313#7194#17#887#242#427#66#1030#5
Frontier modelsmarked frontier in Epoch AI's database0#815#23#533#13#55#33#50#80#84#4
Open weightsshare of its notable models since 2023100%10 of 104%2 of 570%0 of 2214%8 of 5711%1 of 973%19 of 2633%2 of 675%24 of 3267%4 of 683%15 of 18
Largest estimated training cost2023 US dollars, among models Epoch has costed$7M#8DeepSeek-R1 (May 2025)$366M#2GPT-4.5$26M#5Claude 3.5 Sonnet$31M#4Gemini 1.0 Ultra$388M#1Grok 4$53M#3Llama 3.1-405B$4M#9Megatron-Turing NLG 530B$109K#10PLUG$14M#7Mistral Large$21M#6Nemotron-4 340B
Largest training clusterchips for one model, where Epoch records it2,048#7DeepSeek-V3 (Mar 2025)100,000#2GPT-6 Astra—57,000#3Gemini 1.0 Ultra200,000#1Grok 432,000#4Llama 4 Behemoth (preview)4,480#6Megatron-Turing NLG 530B512#8CogView—6,144#5Nemotron-4 340B
Charts on AI Adoptionthat name the company4332223840714

Every chart that names DeepSeek

All charts →
4 charts across 2 hubs name DeepSeek or one of its models. Each has its source and a CSV of its data.
AI Benchmarks2 charts
  1. 205GPT-5 solved 205 PlanBench tasks across eight domains, level with the classical planner LAMA's 204
  2. 93Disguising PlanBench tasks cut DeepSeek R1's solved total from 157 to 93; LAMA still solved 204
AI Infrastructure2 charts
  1. 23.2 WhThe most energy-hungry 2025 models used up to 23 Wh per medium-length prompt
  2. 597 tTraining DeepSeek v3 emitted a fifteenth of the CO₂ of Llama 3.1 405B, despite more parameters

The other companies, same lens

AI Companies →
OpenAI168.8Epoch Capabilities Index65 ledger entriesOpen tracker →Anthropic168.8Epoch Capabilities Index61 ledger entriesOpen tracker →Google166.4Epoch Capabilities Index48 ledger entriesOpen tracker →xAI156.6Epoch Capabilities Index37 ledger entriesOpen tracker →Meta161.8Epoch Capabilities Index33 ledger entriesOpen tracker →Microsoft130.5Epoch Capabilities Index4 ledger entriesOpen tracker →Alibaba157.4Epoch Capabilities Index33 ledger entriesOpen tracker →Mistral AI150.0Epoch Capabilities Index26 ledger entriesOpen tracker →Nvidia146.0Epoch Capabilities Index7 ledger entriesOpen tracker →

How the ledger works

  1. One lens for everyone. Six lenses, same order, same definitions, for every company. No company gets an extra lens for a product it wants to talk about.
  2. Changes, not news. An entry must move a number or a rank on a lens. Launch events, interviews and hires do not count unless they do.
  3. Evidence tier on every line. Independent measure, independent estimate or company data. A company's own claims are replaced when an independent score arrives.
  4. Gaps stay visible. If no source measures a lens for a company, the page says so.
Disclosure: AI Adoption uses AI tools in its research and editorial work, including products from some of the companies tracked here. No company has any say in its page or its ledger, and the same rules apply to every company. Editorial standards · Data sources
Copied with its source. Want the next number like this in your inbox?Get Friday's numbers