DeepSeek logo

DeepSeek

Private (controlled by founder Liang Wenfeng; preparing a mainland China IPO)Founded 2023🇨🇳Hangzhou, Zhejiang, China
CEO

Liang Wenfeng

Data verified as of Sep 4, 2026

Summary

DeepSeek is a Hangzhou-based AI lab, founded in July 2023 by hedge-fund billionaire Liang Wenfeng as a spin-off of his quant fund High-Flyer, that builds frontier open-weight language models at radically low cost. Its January 2025 R1 reasoning model - released under an MIT license and trained for a small fraction of US-lab budgets - briefly made its free chatbot the top US iOS app and triggered a historic market rout in which Nvidia shed roughly $600B in a day. As of mid-2026 its flagship is the DeepSeek-V4 family (V4-Pro at 1.6T total parameters and V4-Flash), open-weight models with 1M-token context reportedly trained on Huawei Ascend chips without Nvidia GPUs. DeepSeek serves ~130M monthly active users in China, though QuestMobile data now ranks it third among domestic AI apps behind ByteDance's Doubao and Alibaba's Qwen, and monetizes through an API priced orders of magnitude below US frontier labs, with annualized revenue reportedly nearing $500M by mid-2026. After operating for three years on High-Flyer's money alone, it closed a record ~$7B maiden round at a ~$52B valuation in June 2026 - with most investors taking non-voting stakes that preserve Liang's control - and is preparing a mainland IPO targeted for 2027, though a planned follow-on raise was paused in late July 2026.

Company OverviewRead the long-form profile of DeepSeek

/ Recent News

View all →
  • View all news

    Browse the daily archive

Main Products

DeepSeek-V4 model family

DeepSeek-V4 model family

Generally Available

DeepSeek's flagship open-weight models: V4-Pro (1.6T total / 49B active parameters) and V4-Flash (284B / 13B active), both MIT-licensed with 1M-token context and hybrid thinking/non-thinking modes - reportedly trained on Huawei Ascend chips without Nvidia GPUs.

Released in preview on April 24, 2026 with weights on Hugging Face. DeepSeek moved a retrained V4-Flash-0731 API into public beta on July 31, 2026, then shipped V4-Pro-0813 to full general availability across app, web and API on August 12, 2026, with new peak/off-peak API pricing taking effect August 16, 2026. Legacy deepseek-chat and deepseek-reasoner endpoints retired July 24, 2026.

V4-Pro Parameters1.6T total / 49B active
Context Window1M tokens
LicenseMIT (open weights)

DeepSeek chatbot (app & web)

Generally Available

Free consumer assistant across iOS, Android and web, running the V4 family with Expert and Instant modes. Briefly the #1 free app on the US iOS App Store after R1's January 2025 launch and still one of the world's top AI chat platforms.

~130M monthly active users in China and ~173M cumulative downloads (estimates); third-largest domestic AI-native app by MAU per QuestMobile (May 2026), behind ByteDance's Doubao and Alibaba's Qwen.

Monthly Active Users~130M in China (QuestMobile, May 2026)
Cumulative Downloads~173M (est.)

DeepSeek API platform

Generally Available

Developer platform exposing the V4 family through OpenAI- and Anthropic-compatible endpoints at aggressive prices (V4-Flash from $0.14/$0.28 per million input/output tokens), with cache-hit input discounts and, from the official V4 release, peak/off-peak pricing.

V4-Flash-0731 is in public beta with Responses-format support for coding tools; V4-Pro-0813 shipped to full general availability on August 12, 2026. New peak/off-peak API pricing takes effect August 16, 2026. Legacy deepseek-chat and deepseek-reasoner models retired July 24, 2026.

V4-Flash Pricing$0.14 / input MTok, $0.28 / output MTok
V4-Pro Pricing$0.435 input / $0.87 output per MTok at launch; rising to $3.96 output at peak hours (half that off-peak) effective August 16, 2026

What's Next

Build out a gigawatt-scale data center in Inner Mongolia

Deploy proceeds of the ~$7B round into self-operated data centers built on Huawei Ascend chips to serve AI-agent workloads, deepening the domestic-silicon strategy proven by V4's Nvidia-free training run. Bloomberg reported on July 30, 2026 that DeepSeek is developing a roughly one-gigawatt facility in Ulanqab, Inner Mongolia, building part of it itself while leasing additional capacity from other providers, and targeting bringing at least part of the capacity online in late 2027 or early 2028. Bloomberg reported on September 4, 2026 that DeepSeek plans to deploy at least 160,000 of Huawei's next-generation Ascend 950DT chips at the facility for running (not training) its models, which would be among the largest known clusters of Chinese AI accelerators; Huawei's limited production capacity is expected to cap 950DT output in the low hundreds of thousands of units this year, so delivery could take more than a year, and DeepSeek has reportedly asked Beijing to help secure faster, larger allocations.

2026-2028

File for a mainland China IPO

Submit the domestic listing filing reported to be targeted for late 2026 toward a public debut on Shanghai's STAR Market in 2027. The follow-on raise paused in late July after founder Liang Wenfeng's leaked investor remarks went viral resumed around August 6, 2026, and by August 29, 2026 DeepSeek was finalizing the round at roughly 50B yuan (~$7.4B) and a ~$74B pre-money valuation with returning backers Monolith Management, Shixiang Capital, Tencent and CATL, expected to close by end of August 2026 with proceeds earmarked for compute buildout and model R&D ahead of the IPO push; no IPO filing has been announced as of late August 2026.

Late 2026 (filing); 2027 (listing)

Build out the Unitree AI pact into shipped embodied-intelligence work

Follow through on the August 6, 2026 agreement disclosed alongside DeepSeek's 140.8M yuan (~$20.8M) strategic stake in Unitree Robotics' Shanghai IPO: a pact to jointly develop AI models for humanoid robots, with Unitree giving DeepSeek priority on model-training procurement and DeepSeek favoring Unitree for robots and embodied-AI applications. It is DeepSeek's first disclosed move into robotics, and the shares carry a 36-month lock-up. No shipped joint output has been disclosed yet.

2026 and beyond

Track Record

1 of 2 announced dates hit · 1 slipped

Ship the official DeepSeek-V4 release

Target Aug 2026 · Resolved Aug 12, 2026

DeepSeek shipped V4-Pro-0813 to general availability across app, web and API on August 12, 2026, graduating it out of preview with a 1 million-token context window, up to 384,000-token output and a focus on agentic tool-use and multi-step workflows. New peak/off-peak API pricing took effect August 16, 2026, raising V4-Pro's peak output price to $3.96 per million tokens from the prior flat $0.87 rate.

Hit

Ship the official DeepSeek-V4 release

Target Jul 2026 · Resolved Aug 1, 2026

The mid-July 2026 target for a full V4 GA passed without one. On July 31, 2026 DeepSeek put only a retrained V4-Flash-0731 API into public beta (outperforming the V4-Pro preview on nine published agent/coding benchmarks), while the V4-Pro API and the app/web models were unchanged. Chinese tech media reported the complete rollout, including peak/off-peak pricing, sliding to a possible August 10-20, 2026 window tied to closed testing of DeepSeek's new 'Harness' agentic coding tool.

Slipped

Operations & Revenue

StatusScaling; IPO preparation

DeepSeek runs one of China's most used AI chatbots (~130M MAU, though QuestMobile now ranks it third domestically behind Doubao and Qwen) and the leading open-weight model family, with annualized revenue nearing $500M and a headcount push doubling across departments from a deliberately small base. Freshly capitalized by a record ~$7B June 2026 round, it is building its own data centers on domestic Huawei Ascend compute for AI-agent workloads and preparing a mainland China IPO filing for a potential 2027 listing. A planned follow-on raise, paused in late July 2026 after leaked founder remarks, reopened around Aug 6, 2026 targeting ~$8B at a ~$74B valuation, not yet closed. On August 13, 2026 it also open-sourced DeepSeek Harness, an agentic coding runtime positioned against Anthropic's Claude Code, alongside V4-Pro's move to general availability.

Revenue Streams

API platform

Usage-based, per-token access to the V4 family via OpenAI- and Anthropic-compatible endpoints, priced orders of magnitude below US frontier labs; peak/off-peak pricing arrives with the official V4 release.

Consumer chatbot (free)

The DeepSeek app and chat.deepseek.com are free - a deliberate user-growth strategy that leaves the API as the primary revenue source for now.

Key Metrics

Employees

~150-200 (mid-2026); doubling headcount across departments underway

Est. Annual Revenue

Annualized run-rate reportedly nearing $400-500M by mid-2026, roughly double the ~$200-220M estimated for 2025 - monetization is API-led, with the consumer chatbot free

Valuation

~$52B (Jun 2026); a follow-on round paused Jul 25, 2026 after founder remarks leaked reopened around Aug 6, 2026 and by Aug 29, 2026 was nearing close at ~50B yuan (~$7.4B) raised and a ~$74B pre-money valuation, expected to close by end of August 2026

Monthly Active Users

~130M in China (QuestMobile, May 2026, published Jul 14 - the most recent published QuestMobile snapshot as of late August 2026); third-largest Chinese AI-native app behind ByteDance's Doubao (~382M MAU) and Alibaba's Qwen (~167M MAU)

China Chatbot Market Share

Third among China's AI-native apps by MAU (QuestMobile, May 2026), behind Doubao and Qwen - the earlier ~89% 'dominant' estimate reflected the Jan 2025 R1-launch spike and no longer holds

App Downloads

~173M cumulative since Jan 2025 (est.)

V4 API Pricing

V4-Flash $0.14/$0.28 per MTok; V4-Pro $0.435 input/$0.87 output per MTok at launch, rising to $3.96 output at peak hours (half that off-peak) effective August 16, 2026 following V4-Pro's August 12, 2026 GA release

Claimed V3 Training Cost

~$5.6M (2.788M H800 GPU-hours, per DeepSeek's own Dec 2024 V3 technical report) vs ~$100M+ est. for GPT-4

Employees

~150-200 (mid-2026); DeepSeek announced Jun 25, 2026 plans to double headcount across departments (pre-training data, server engineering, AI search, data-center ops, a new Harness agent team) following its $7.4B raise

Timeline

2026Anthropic alleges industrial-scale distillation

In February 2026 Anthropic publicly accuses DeepSeek - along with Moonshot AI and MiniMax - of harvesting Claude outputs for model distillation through roughly 24,000 fraudulent accounts and 16M+ exchanges, sharpening the US–China AI dispute; DeepSeek is also barred from US defense and intelligence systems under the FY2026 NDAA.

2026DeepSeek-V4 family debuts, trained on Huawei silicon

On April 24, 2026 DeepSeek releases previews of V4-Pro (1.6T total / 49B active parameters) and V4-Flash (284B / 13B active) - MIT-licensed open-weight models with 1M-token context, reportedly trained on Huawei Ascend chips with no Nvidia GPUs, a landmark for China's domestic-compute push.

2026Record ~$7B maiden funding round at ~$52B valuation

In June 2026 DeepSeek closes its first-ever external round - roughly ¥51B (~$7B), the largest single AI funding round in China's history - at a ~$52B post-money valuation, with Tencent, CATL, JD.com, NetEase and a state AI fund participating, mostly via non-voting stakes that keep Liang Wenfeng in control.

2026IPO preparation and talks for a further ~$1.5B

In mid-July 2026 Bloomberg and the FT report DeepSeek is in talks to raise a further ~$1.5B at a ~$71B valuation and is preparing a mainland China IPO filing this year for a potential 2027 listing.

2026V4-Flash-0731 update posts major agentic gains

On July 31, 2026 DeepSeek publishes DeepSeek-V4-Flash-0731, a re-post-trained update to its 284B-total/13B-active-parameter model that posts large agentic and coding benchmark gains over its own V4-Pro despite far fewer active parameters, and moves the V4-Flash API into public beta with support for the Responses format used by coding tools like Codex.

2026Takes strategic stake in Unitree, signs embodied-AI pact

DeepSeek's parent entity becomes a strategic placement investor in Unitree Robotics' Shanghai STAR Market IPO, buying 933,400 shares (~141M yuan, ~$20.8M, under a three-year lock-up) alongside Tencent and several state-backed institutions. The two companies agree to co-develop AI models and embodied-intelligence technology and to give each other priority access to robot procurement and AI model services - DeepSeek's first disclosed move into humanoid robotics.

2026Follow-on funding round resumes at ~$74B valuation

DeepSeek reopens its paused follow-on funding round around August 6, 2026, seeking close to $8B at a valuation near 500B yuan (~$74B), with Monolith Management - an early Moonshot AI backer - reportedly in talks to participate. The round had been paused since July 25 after leaked remarks from founder Liang Wenfeng frustrated some investors.

2026V4-Pro reaches general availability

On August 12, 2026 DeepSeek ships V4-Pro-0813 to general availability across its app, web interface and API, exiting preview with a 1 million-token context window, up to 384,000-token output and a focus on agentic tool-use and multi-step workflows. New peak/off-peak API pricing takes effect August 16, 2026, raising V4-Pro's peak output price to $3.96 per million tokens from the prior flat $0.87 rate.

2026Ships DeepSeek-V4-Flash-Vision-Exp, its first multimodal model

On August 21, 2026 DeepSeek releases DeepSeek-V4-Flash-Vision-Exp, an experimental model that opens its API to image inputs alongside text for the first time, billed at the same per-token rate as V4-Flash. On the Agents Last Exam benchmark it scores 27.3 versus 25.7 for Anthropic's Opus 4.8, and DeepSeek says its vision pipeline uses roughly 90 KV-cache entries per image versus about 870 for Claude's vision system.

2026Open-sources DeepSeek Harness, an agentic coding runtime rivaling Claude Code

On August 13, 2026, alongside V4-Pro's general-availability launch, DeepSeek releases DeepSeek Harness v0.1 as open-source software under an MIT license - a plugin-first agent runtime for assembling coding agents from replaceable models, tools, skills, sessions, and sandboxes, positioned as a direct alternative to Anthropic's Claude Code. The developer preview passes roughly 95,000 GitHub stars within about two days of release.

2025R1 release triggers a $600B Nvidia rout

On January 20, 2025 DeepSeek releases the MIT-licensed R1 reasoning model. Within a week its free app tops the US iOS App Store and, on January 27, Nvidia falls ~18% - erasing roughly $600B, the largest single-company one-day market-value loss in US history - in what is dubbed AI's 'Sputnik moment.'

2025V3.1 and V3.2 fold reasoning into hybrid models

DeepSeek ships V3.1 in August 2025 (hybrid thinking/non-thinking modes) and V3.2 with its Sparse Attention architecture on December 1, 2025 - absorbing the R-series reasoning line into the main model family after the long-rumored R2 never ships.

2024DeepSeek-V2 ignites China's LLM price war

DeepSeek releases V2 in May 2024 at disruptively low API prices, earning the nickname 'the Pinduoduo of AI' and forcing Chinese rivals including Alibaba, Baidu and Tencent to slash their own model pricing.

2024DeepSeek-V3 trained for a claimed ~$6M

In December 2024 DeepSeek releases V3, a frontier-class mixture-of-experts model it says was trained for roughly $6M - versus an estimated ~$100M for GPT-4 - establishing its extreme cost-efficiency playbook.

2023Founded as a High-Flyer spin-off

Liang Wenfeng founds DeepSeek in Hangzhou on July 17, 2023, spun out of and wholly funded by his quantitative hedge fund High-Flyer, with a deliberately small research-first team.

Funding

RoundDateAmountInvestorsSource
Internal funding (High-Flyer)2023–2026Undisclosed (wholly funded by High-Flyer)High-Flyer / Liang Wenfeng (84% founder stake as of May 2024)
Maiden external roundJun 2026~$7B at ~$52B valuationTencent, CATL, JD.com, NetEase, Monolith, Loyal Valley Capital, China's National AI Industry Investment Fund; ¥20B from Liang Wenfeng
Follow-on (resumed)Jul-Aug 2026~50B yuan (~$7.4B) at a ~$74B pre-money valuation; DeepSeek paused signing of new agreements on Jul 25, 2026 after founder Liang Wenfeng's leaked investor remarks went viral, then reopened the round around Aug 6, 2026 and by Aug 29, 2026 was finalizing the raise with returning backers Monolith Management, Shixiang Capital, Tencent and CATL, expected to close by end of August 2026Monolith Management, Shixiang Capital, Tencent, CATL

/ Related Companies

All AI Labs
Anthropic logo

Anthropic

AI Labs

Anthropic is a San Francisco AI-safety and research company, founded in 2021 by former OpenAI researchers led by siblings Dario and Daniela Amodei, that develops the Claude family of large language models. As of mid-August 2026 its lineup spans Claude Haiku 4.5, Claude Sonnet 5, the everyday-default Claude Opus 5 (which replaced Opus 4.8 on July 24), and its most capable widely released model, Claude Fable 5 (plus the invitation-only Claude Mythos 5). Claude reaches customers through the claude.ai apps, the Claude API, enterprise and government plans, and Claude Code - its agentic coding tool, which runs at a $2.5B+ run-rate with 2M+ weekly active users - alongside the Claude Cowork 'AI coworker' and the June 2026 Claude Science research workbench, built on the ~$400M acqui-hire of ex-Genentech drug-discovery startup Coefficient Bio, that positions Anthropic in AI-driven pharma R&D. Anthropic differentiates on safety and interpretability research and runs one of the industry's largest multi-vendor compute footprints spanning Amazon (Project Rainier; Amazon's committed capital reached $33B in April 2026), Google (TPUs, up to $40B plus a 3.5 GW Broadcom-built expansion), AMD, CoreWeave, and newer data-center leases with TeraWulf, Riot Platforms and Volta Infra - and was reported on August 13, 2026 to be in talks to acquire AI-infrastructure startup Decart for roughly $6B. Annualized run-rate revenue topped $65B as of late July 2026 (up from ~$47B in May and ~$9B at the end of 2025; some IPO investors are modeling $100-120B by year-end), more than 1,000 customers now spend over $1M/year, and preliminary Q2 2026 revenue exceeded $11.5B with its first profitable quarter since founding. In May 2026 it raised a $65B Series H at a $965B valuation (secondary markets have since priced it above $1T-$1.2T) and confidentially filed for an IPO in June 2026; CFO Krishna Rao is now holding early investor meetings for a possible fall listing that some backers are modeling near a $2T valuation, though Anthropic itself has not set a target. On August 27, 2026 a San Francisco federal judge (Rita Lin) ruled the Pentagon's supply-chain-risk designation of Anthropic unlawful as First Amendment retaliation and arbitrary and capricious, writing that the measures were illegal and baseless; a separate D.C. Circuit case over a second designation remains pending, and the government is expected to appeal. The same day Anthropic opened a research preview of the Model Hardware Standard (MHS), a model-agnostic specification started with HHMI Janelia that lets AI agents operate lab and manufacturing hardware such as microscopes, liquid handlers and robotic arms, with a plan to open-source the standard after the preview.

Apptronik logo

Apptronik

AI Labs

Apptronik builds Apollo, a 5 foot 8 inch, 160 pound humanoid that lifts boxes up to 55 pounds, and it is the humanoid company Google DeepMind chose as its hardware partner. Spun out of the University of Texas at Austin's Human Centered Robotics Lab in 2016 on the back of roughly 15 earlier robots including NASA's Valkyrie, it unveiled Apollo in 2023 and now has units working in designated areas inside factories and warehouses run by Mercedes-Benz, GXO Logistics, and contract manufacturer Jabil. Funding came in two waves: a $350 million Series A in February 2025 co-led by B Capital and Capital Factory with Google participating, later extended, and then a $520 million Series A-X extension on February 11, 2026 that took the Series A past $935 million and total capital to nearly $1 billion at a valuation around $5 billion. In June 2026 it unveiled Apollo 2, available in both bipedal and wheeled-base configurations, and opened Robot Park, a roughly 90,000 square foot data-collection and training facility in Austin. When Google DeepMind launched Gemini Robotics 2 on July 30, 2026, the whole-body-control demonstration ran on Apollo 2 rather than on Boston Dynamics' Atlas. The honest state of the business is that Apollo is still in pilots: Apptronik has never published a price, and CEO Jeff Cardenas has pointed at 2027, with the commercial Apollo 3, as the year volume orders are supposed to arrive.

Aurora Innovation logo

Aurora Innovation

AI Labs

Aurora Innovation is a self-driving vehicle company now focused on commercial autonomous trucking, with the Aurora Driver deployed first as Aurora Driver for Freight. After beginning regular driverless customer deliveries between Dallas and Houston in late April 2025, Aurora has tripled its driverless network to 10 routes across the Sun Belt - including a roughly 1,000-mile Fort Worth–Phoenix lane - and in May 2026 ran its first lane outside Texas, a ~200-mile Dallas–Oklahoma City route with Volvo's VNL Autonomous. By June 2026 it had surpassed 440,000 driverless miles with zero Aurora Driver-attributed collisions and 100% on-time performance. Customers now include Werner, Hirschbach, McLane (Berkshire Hathaway), Detmar, Uber Freight, Value Truck, Charger Logistics, and others, and Aurora runs both an asset-light Driver-as-a-Service subscription model (carrier owns the trucks) and a Transportation-as-a-Service model (Aurora owns and operates the trucks), with vehicle partners PACCAR, Volvo, and AUMOVIO. Q2 2026 revenue was $2M against a $270M net loss, and management reaffirmed guidance of $14-16M for 2026 with a goal of exiting the year with 200+ driverless trucks (~20-25 by end of Q3) and ~$80M run-rate.

Baidu Apollo logo

Baidu Apollo

AI Labs

Baidu Apollo is Baidu's autonomous-driving platform and the operator of Apollo Go, one of the world's largest fully driverless robotaxi networks. On its August 18, 2026 Q2 earnings call Baidu disclosed Apollo Go has delivered more than 23 million cumulative public rides and over 350 million autonomous kilometers across a 28-city global footprint (including a first Asian expansion beyond China and Hong Kong, into South Korea, in February 2026), with roughly 1 million fully driverless rides in the quarter alone and a safety record of one airbag deployment per 14.4 million km, with management saying Apollo Go has reached unit-economics breakeven in its largest China operating city. That international growth came alongside a rockier domestic quarter: a March 31, 2026 mass system failure stranded over 100 Apollo Go vehicles on elevated highways in Wuhan (no injuries), prompting Chinese regulators to pause new robotaxi permits nationwide for several months; Baidu said the resulting operational adjustments weighed on ride volume in certain domestic cities before operations began resuming on a stronger footing in August. The same Q2 quarter, China's Ministry of Industry and Information Technology issued GB 44721-2026, the country's first mandatory national safety standard for L3/L4 automated driving, taking effect July 1, 2027. Apollo Go runs fully driverless commercial service in China, Dubai, and on Abu Dhabi's Yas Island, began Level 4 open-road trials in eastern Switzerland (AmiGo, with PostBus) on June 1, 2026, and has kept widening its Hong Kong testing scope - an August 14 approval added the Southern District to its North Lantau driverless zone, its third expansion there in about six months, pushing cumulative safe driverless testing distance in the city past 20,000 km. It also runs safety-operator road testing in London with Freenow/Lyft and Uber, begun July 28, and in July 2026 signed an MoU with Kazakhstan's Turlov Private Holding to explore autonomous mobility services, the first move by a Chinese robotaxi operator into Central Asia.

Google DeepMind logo

Google DeepMind

AI Labs

Google DeepMind is Alphabet's consolidated artificial-intelligence research lab, formed in April 2023 by merging the original DeepMind (founded in London in 2010, acquired by Google in 2014) with Google Brain. Co-founder and Nobel laureate Demis Hassabis stepped down as CEO on August 5, 2026, moving to Chair of Google DeepMind and Chief Scientist of Alphabet to focus on long-term AGI research and Isomorphic Labs; Koray Kavukcuoglu, an SVP reporting directly to Sundar Pichai, now runs day-to-day operations and the Gemini roadmap. The lab builds the Gemini family of frontier multimodal models - which powers the Gemini app (900M+ monthly active users as of mid-2026), Google Search's AI Mode and AI Overviews, Google Cloud's Vertex AI, and, via a 2026 partnership, Apple's next-generation Siri. Beyond consumer AI, DeepMind pioneered game-playing systems (AlphaGo, AlphaZero) and scientific breakthroughs including AlphaFold - whose protein-structure database has been used by 3M+ researchers and earned Hassabis and John Jumper the 2024 Nobel Prize in Chemistry - plus AlphaGenome, AlphaEvolve, Veo and Genie world models, and Gemini Robotics. Its drug-discovery spinout Isomorphic Labs is preparing first-in-human oncology trials. The August leadership change follows a stretch of talent departures (Jumper among them) and delays shipping Gemini 3.5 Pro, and coincided with Google chief scientist Jeff Dean also leaving the company after 27 years.

Mistral AI logo

Mistral AI

AI Labs

Mistral AI is Europe's leading frontier AI lab, founded in April 2023 in Paris by CEO Arthur Mensch (ex-Google DeepMind) with chief scientist Guillaume Lample and CTO Timothée Lacroix (both ex-Meta). It champions open-weight models: its December 2025 Mistral 3 generation includes Mistral Large 3, a 675B-parameter sparse mixture-of-experts model released under Apache 2.0, alongside commercial and specialist models (Medium 3.5, Small 4, Magistral reasoning, Codestral/Devstral coding, Voxtral audio, OCR 4 documents and the new Robostral robotics line). Its consumer assistant Le Chat was rebranded 'Vibe' in May 2026 and repositioned as an agentic work-and-code platform. Positioned as Europe's sovereign alternative to US labs, Mistral has deep industrial and government ties - lead Series C investor ASML (~11% stake), Airbus, BMW, Stellantis, CMA CGM (a €100M deal) and a framework agreement with France's Ministry of Armed Forces - and is building its own Nvidia-powered infrastructure through Mistral Compute, whose first 44 MW data center south of Paris came online in mid-2026, alongside a €1.2B second site under construction with EcoDataCenter in Borlänge, Sweden - Mistral's first infrastructure investment outside France, targeted to open in 2027. Annual recurring revenue passed $400M in early 2026, up roughly 20x year over year, with a $1B+ target by year-end. In June-July 2026 the company was reported to be in talks to raise about €3B at a ~€20B valuation, with EQT's new Scaleup Europe Fund reported to lead or co-lead; as of September 2, 2026 no primary source has confirmed the round closed. Enterprise reach extends beyond its core industrial deals to UK grocer Tesco (three-year deal and joint AI lab, December 2025) and a Southeast Asia push anchored in Singapore, where headcount is set to triple to 100 by end-2026.