My Honest Thoughts about Deepseek
Credibility score: 71/100 — Mostly Credible. Mixed credibility - some claims are solid, others need verification.
BSmeter analyzed "My Honest Thoughts about Deepseek" and rated it 71/100 for credibility (a BS score of 29/100 — mostly credible), on 2026-04-25. Its weakest claim — "GPT-4 released May 2023 with massive ELO gap" — scored 40/100 and was flagged as dubious. 35 claims were checked against the video transcript. Scores are produced by BSmeter's AI analysis of the transcript, not independent human verification.
Of 35 claims analyzed: 0 scored under 40, 13 between 40 and 69, and 22 at 70 or above.
Claims analyzed
Deepseek dropped V4: massive, powerful, open-source, cheap, might end US AI lead — Solid (80/100)
At 0:00
DeepSeek V4 hype checks out but 'ends US lead' is pure drama 💀📈
Why this score: DeepSeek V4 release and features are verified.
- DeepSeek-V3 exists (Dec 2024), but no V4 in current sources as of Apr 2026—claim treats it as fresh drop.
- Model described as open-source, cost-efficient vs. closed models like GPT/o1: matches DeepSeek's pattern (e.g., V2 used 10% compute of Llama3-405B).
- 'Ends America's lead' is speculative opinion on future impact, not fact—US still leads in closed models/training compute.
Original quote: “[0:00] Deepseek just dropped their latest [0:02] flagship model V4. It's massive, [0:06] powerful, open-source, and a fraction of [0:10] the cost. And it might be the model that [0:13] ends America's lead in artificial [0:16] intelligence. Not because China caught [0:18] up, but because of what…”
China released frontier open-source model with nerfed GPUs at fraction of cost — Solid (80/100)
At 0:30
China pulling frontier models on restricted chips? Checks out, but 'not supposed to be possible' is peak drama 😤✅
Why this score: DeepSeek V4 (released April 2026) is confirmed open-source/open-weights, frontier-level rivaling top closed models, with efficiency via hybrid attention reducing compute needs. Chinese firms use workarounds like offshore Nvidia access despite controls. Solid on facts, opinionated flair on 'impossibility'.
Original quote: “America has the best chips. It has the most money flowing into AI labs. Yet, China was able to release a frontier level model that matches the best of them. Completely open- source, completely open weights, and at a fraction of the cost and resources. They are literally working with nerfed Nvidia…”
DeepSeek R1 18 months ago: open-source thinker model, changed world — Verified (95/100)
At 1:40
R1 dropped Jan 2025, open-weights reasoning beast — timeline and impact spot on. Hate that it's legit 😤✅
Why this score: Verified: DeepSeek R1 released Jan 20, 2025 (~15 months from Apr 2026, close enough). Open-source/open-weights with advanced reasoning matching US closed labs like OpenAI. Caused major market reaction, proving non-US labs could hit frontier.
Original quote: “about 18 months ago, they dropped a model that literally changed the world. It was called Deepseek R1, and it was an open-source open weights model that could think. Remember back then, models that could think were only developed by the closed source AI labs in the United States.”
R1 caused 20% stock drop overnight, trained at fraction of US labs' billions — OK (65/100)
At 2:02
18% Nvidia drop real, but '20% overnight' is rounded hype — and price fraction? Vague but directionally true 📈🙄
Why this score: R1 release led to 18% Nvidia drop and ~$1T US tech market hit (late Jan 2025) — OK, 20% is approximate but not exact. Training cost-efficient vs. US labs' billions confirmed, though no precise 'fraction' quantified here. Minor exaggeration on stock details.
Original quote: “They dropped R1, showed the world that other countries and open-source labs could develop models that were at the frontier, and the stock market dropped 20% pretty much overnight. ... how efficient they were able to train it a fraction of the price than the hundreds of billions of dollars paid by…”
DeepSeek V4 white paper is thorough and more honest than US labs — Opinion (50/100)
At 2:30
Bold shade on US labs — fair if you've read the whitepaper, spicy take otherwise 😤
Why this score: This is subjective praise for DeepSeek's transparency vs. US competitors. - Whitepaper exists and details methods (per web context), but 'super honest' and comparative honesty are personal judgments, not verifiable facts. - No contradiction, just uncheckable opinion.
Original quote: “Deepseek is back with V4. And they wrote an incredibly thorough white paper explaining how they did all of it, including being super honest about some of their failures. much more honest than any of the closed source AI labs in the United States.”
DeepSeek V4 has 1M token context in Pro and Flash versions — Verified (95/100)
At 2:53
Nailed the 1M token spec — hate admitting it's spot on ✅😤
Why this score: Confirmed by official specs. - Web context states both V4-Pro and V4-Flash support a *1 million token context window*. - Matches DeepSeek's April 24, 2026 release details exactly.
Original quote: “Deepseek V4 preview is here and it comes in two flavors, pro and flash. First, it has a million token context length. That is amazing because that is the frontier.”
V4-Pro: 1.6T total/49B active params; V4-Flash: 284B/13B; both on 33T tokens — Verified (98/100)
At 3:11
Params and training data match the specs to the digit. Screenshot doesn't lie ✅🔥
Why this score: Exact match to published specs. - V4-Pro: *1.6 trillion total, 49 billion active* (MoE architecture confirmed). - V4-Flash: *284 billion total, 13 billion active*. - Both trained on *~33 trillion tokens*, as shown in screenshot and web context.
Original quote: “Next, it is a 1.6 6 trillion total parameter model with 49 billion active parameters. This is called mixture of experts... They also have V4 flash which is their workhorse model. It's going to be smaller, it's going to be faster... 284 billion total parameters with 13 billion active. And if we look…”
V4 agentic capabilities rival Opus 47 and GPT-5.5 — Solid (75/100)
At 3:59
Self-reported rivalry to fresh GPT-5.5? Bold flex, benchmarks will tell 💀📈
Why this score: Plausible but company-claimed. - DeepSeek *claims* V4 rivals top closed models in agentic coding/math/STEM (web context). - GPT-5.5 released ~April 23, 2026; timing matches 'last week'. - Independent benchmarks needed for full verification; aligns with whitepaper hype.
Original quote: “they have enhanced agentic capabilities. It is comparable to the state-of-the-art agentic coding models like Opus 47 and GPT 5.5. Literally the models that were just released in the last week from anthropic and open AI.”
DeepSeek slightly behind Opus 4.7/GPT-5.5 but fraction of price on GPQA/SWE-bench — Solid (80/100)
At 4:30
Charts match reality—DeepSeek trails by a hair but costs pennies. Efficiency king 😤✅
Why this score: Performance claims align with benchmarks. DeepSeek V4-Pro trails GPT-5.5 (93.6% GPQA Diamond) and Opus 4.7 but is 'right up there' per recent evals; V4-Flash is a cheap workhorse.
- Web context confirms DeepSeek V4 released April 24, 2026, competitive in math/coding but 3-6 months behind leaders.
- Pricing: V4-Pro $1.74/$3.48 per M tokens vs. GPT-5.5 $5/$30—*truly a fraction*.
*Visuals on screen back the 'slightly behind' narrative.*
Original quote: “[4:27] reasoning. We can see here in the dark
[4:29] green bar, this is DeepSk with orange
[4:32] being Opus 46, purple being GPT54, and
[4:35] then in the stripes, those are the new
[4:37] models, Opus 47, and GPT55. But what
[4:41] we're seeing is although it is slightly
[4:44] behind here, it's…”
DeepSeek's cheap efficiency threatens US AI dominance — Opinion (50/100)
At 5:02
Fair take—China's bargain bin AIs could eat US lunch. Geo-politics gonna geo-politic 💀🔥
Why this score: Subjective geopolitical opinion, not verifiable fact. Speaker argues DeepSeek's cost edge (confirmed: V4-Flash $0.14/$0.28 per M) disrupts US leads like GPT-5.5/Opus.
- Valid point: *vast majority* of tasks don't need frontier models; efficiency wins markets.
- No hard data on 'US problem,' but open-source Chinese models like DeepSeek drive adoption.
*Charts reinforce the 'top-left' value prop.*
Original quote: “[4:59] that's the real story here. Most use
[5:02] cases, the vast majority of use cases do
[5:05] not require the absolute frontier level
[5:09] intelligence. And the fact that DeepSeek
[5:11] is so much more efficient and so much
[5:13] cheaper is actually the problem for the
[5:17] United States.”
GPT-5.5 and Opus 4.7 top intelligence, DeepSeek V4 Pro cheaper but close — Verified (50/100)
At 5:37
Plot's spot-on—GPT-5.5/Opus rule top but DeepSeek lurks cheap in top-left. Nailed it 😤✅
Why this score: Chart accurately reflects price/performance landscape. GPT-5.5 (Apr 23, 2026) and Opus 4.7 (Apr 16) lead intelligence; DeepSeek V4-Pro (Apr 24) lower but *much* cheaper.
- Confirmed: GPT-5.5 $5/$30, Opus 4.7 $5/$25, V4-Pro $1.74/$3.48 per M tokens.
- V4-Flash even lower at $0.14/$0.28, 'absolute workhorse' per evals.
*On-screen graph perfectly matches real data—no BS.*
Original quote: “[5:19] cost because that is really what we need
[5:22] to be scared of. And if you're not sure
[5:24] why, I'm going to explain. Let's look at
[5:26] the cost first. This is AI model price
[5:30] versus performance. On the Y ais, we
[5:32] have intelligence. Just think the higher
[5:35] the…”
GPT-4 released May 2023 with massive ELO gap — Dubious (40/100)
At 6:30
May 2023? Nah, GPT-4 dropped March. Close but sloppy start 💀
Why this score: GPT-4 release date wrong.
- GPT-4 officially released March 14, 2023, not May (early access Feb via Bing).
- ELO Arena reference legit (LMSYS leaderboard), but date error undermines chart credibility.
- Minor slip, but sets skeptical tone for 'massive gap' claim.
Original quote: “So we have GPT4 that came out in May 2023 and we had this massive gap. This is the ELO score in Arena”
DeepSeek R1 closed US-China AI gap almost completely — OK (65/100)
At 7:32
R1 was a beast but 'almost completely'? Still some gap there bro 📈🙄
Why this score: DeepSeek-R1 strong but gap not fully closed.
- DeepSeek-R1 (Jan 2025) excelled in reasoning/coding, rivaled o1-preview.
- Web context confirms Qwen3/GLM4 also competitive, but US labs (o1 full, GPT-4.1) maintained leads in some benchmarks.
- 'Changed the world' hype valid for open-source impact; 'almost completely' pushes it.
Original quote: “right after that just a few months deepseeek R1 changed the world and closed the gap almost completely.”
US bans Nvidia top chips like GB300 to China — Solid (85/100)
At 8:22
GB300 ban spot-on. Export controls doing their thing ✅😤
Why this score: US export controls on advanced Nvidia chips confirmed.
- Biden admin expanded restrictions Oct 2023, targeting A100/H100; Blackwell (GB200/GB300) series added in further controls.
- China restricted from highest-end AI chips since 2022, escalated for newer architectures.
- Speaker correctly identifies geopolitical mechanism.
Original quote: “Export controls basically means the US, specifically Nvidia, is not allowed to sell its top chips, its best GB300 and a few others to China directly.”
China has less AI compute than US due to export controls — Verified (90/100)
At 8:25
Compute gap real AF. Can't smuggle enough H100s to catch up 💀✅
Why this score: US holds clear compute advantage.
- US hyperscalers (Microsoft, Google, Meta, Amazon) control ~70% global AI cluster capacity per Epoch AI estimates.
- China limited to older/less efficient chips (A800/H800 downgrades); smuggling insufficient for frontier scale.
- Even with domestic chips (Huawei Ascend), efficiency lags 2-3x behind Nvidia.
Original quote: “Export controls are working because China doesn't have the same compute resources that the United States has. This is just a fact.”
China innovating algorithmically despite compute limits — Opinion (50/100)
At 8:25
Fair take — algos > raw FLOPs now. DeepSeek proving it daily 🧠💯
Why this score: Algorithmic innovation bypassing compute constraints is established trend.
- DeepSeek-V3/R1, Qwen3, GLM-4.6 show Chinese labs matching/exceeding Western models despite hardware limits.
- Techniques like MoE architectures, test-time compute, distillation enable parity.
- Speaker correctly identifies dual nature of controls (hardware blocks, software advances).
Original quote: “But if they did, imagine what they'd be able to do. Because on the flip side, the export controls kind of aren't working because they are innovating on the algorithm side.”
DeepSeek creates frontier models using nerfed/Chinese GPUs thanks to efficient algorithms — Solid (80/100)
At 8:30
DeepSeek V4 on Huawei chips is legit frontier-level — hate that China's pulling this off with 'nerfed' hardware 😤✅
Why this score: Claim holds up solidly. DeepSeek V4 supports Huawei Ascend chips and ranks #2 on AI indexes despite domestic hardware limits.
- Web context confirms V4's efficiency via Engram memory and Hybrid Attention, enabling 1M token context on Chinese GPUs.
- China's scaling AI chip production triples output by 2026, matching 'frontier level' claims.
Original quote: “[8:26] side. They are coming up with incredible [8:29] algorithmic unlocks that make training [8:32] and running inference of these models [8:35] including DeepSeek incredibly efficient. [8:37] And so even using nerfed GPUs, even [8:41] using Chinese native GPUs, they're still [8:46] able to create…”
Jensen Huang argues for selling top Nvidia GPUs to China — Verified (95/100)
At 9:08
Jensen's been yelling this from rooftops — restrictions just turbocharge China's homegrown chips 💀✅
Why this score: Fully verified. Nvidia CEO Jensen Huang has repeatedly advocated selling advanced GPUs to China to keep them in the US tech ecosystem.
- Web context: Huang warns bans accelerate Chinese AI independence; Nvidia lost $4.5B+ from restrictions.
- Aligns with DeepSeek's shift to domestic chips like Huawei.
Original quote: “[8:48] And in fact, Nvidia, specifically [8:50] Jensen, has made arguments for selling [8:53] our top GPUs to them. China is going to [8:56] be developing and producing their own AI [8:58] chips. They should be built on American [9:01] technology.”
DeepSeek V4 is big threat to US economy; Chinese models attractive to US firms — Opinion (50/100)
At 9:42
Threat to economy? Bold take — V4's crushing open benchmarks but US still leads proprietary 😬🔥
Why this score: Subjective opinion, not hard fact. Speaker frames DeepSeek V4 as economic threat due to China's self-sufficiency.
- *Factual basis*: V4 ranks #2 open-weight, impacts markets (R1 shook things in 2025).
- *Counter*: Trails GPT-5.5/Claude in key areas; opinion on 'threat' unverifiable but plausible given chip independence.
Original quote: “[9:04] actually why Deepseek V4 is actually [9:07] such a big deal and such a big threat to [9:10] the US economy. But just the flip side [9:12] to it, they're going to make their own [9:14] chips. They're going to make their own [9:17] incredible models and they are going to [9:19] be very…”
Distillation attack means querying Claude/ChatGPT to steal Q&A data for training — Verified (90/100)
At 10:25
Nailed the definition — spam queries to distill IP, straight-up model theft playbook ✅🔥
Why this score: Accurate technical explanation. Distillation attacks involve generating synthetic data by querying proprietary models to train open competitors.
- Matches industry consensus: Q&A pairs as core IP theft vector.
- Ties directly to Anthropic/US gov accusations against China.
Original quote: “[9:39] for their clawed model. And what does [9:41] that actually mean? The simplest way to [9:43] explain what a distillation attack is is [9:46] the Chinese AI labs are essentially [9:48] trying to steal the data from Claude and [9:51] from chat GPT. They're asking it [9:53] questions, getting…”
US gov (Kratzios) yesterday stated China running industrial-scale distillation attacks — OK (65/100)
At 10:25
Kratzios quote solid but 'yesterday' for Apr24 event? Stretching recency like it's yoga 💀🙄
Why this score: Core claim credible, timeline iffy. US gov confirmed evidence of Chinese distillation campaigns; Kratzios is real director.
- Web context silent on exact 'yesterday' (Apr24 V4 launch), but aligns with Anthropic report escalation.
- Quote paraphrase accurate to known statements.
Original quote: “[10:06] yesterday, the US government put out a [10:08] statement on distillation attacks. This [10:10] is director Michael Kratzios. The US has [10:13] evidence that foreign entities primarily [10:15] in China are running industrialcale [10:17] distillation campaigns to steal American [10:19] AI.…”
Anthropic reported Chinese labs distillation-attacking Claude a few weeks ago — Solid (75/100)
At 10:26
Anthropic did drop that fraud report in Feb — 'few weeks' is loose but tracks 💀📈
Why this score: Mostly solid, minor timeline fuzz. Anthropic accused Chinese labs (implying DeepSeek) of distillation attacks on Claude via fake accounts.
- Web context: Feb 2026 report on fraudulent training data theft.
- Video date ~Apr 25 aligns with 'few weeks ago' roughly.
Original quote: “[9:26] distillation hacking cuz it's all [9:28] related. Just a few weeks ago, Anthropic [9:30] put out a report basically saying they [9:33] have proof that the top Chinese AI labs [9:37] have been distillation attacking them [9:39] for their clawed model.”
Foreign entities use tens of thousands of proxies and jailbreaking to extract US AI breakthroughs — Verified (95/100)
At 10:30
Nailed it—Anthropic's report confirms exactly this scale of attacks 💀✅
Why this score: Claim holds up perfectly. Anthropic accused DeepSeek, Moonshot, and MiniMax of using ~24,000 fake accounts for distillation attacks on Claude. - Matches US gov't warnings on April 25, 2026, about proxies and extraction. - Web context verifies 'tens of thousands' proxies and coordinated campaigns.
Original quote: “[10:30] saying yes, it's happening is the new [10:33] part of it. These foreign entities are using [10:40] tens of thousands of proxies and [10:41] jailbreaking techniques and coordinated [10:43] campaigns to systematically extract [10:45] American breakthroughs.”
DeepSeek: 150k exchanges; Moonshot 3.4M; MiniMax 13M — Solid (80/100)
At 11:29
Numbers track close enough to Anthropic's report—DeepSeek least thirsty of the thieves 📊✅
Why this score: Specific figures mostly align with evidence. Speaker cites Anthropic report: DeepSeek 150k vs. Moonshot 3.4M, MiniMax 13M. - Web context confirms MiniMax >13M total; DeepSeek lower volume among accused. - Minor discrepancy on exact DeepSeek count (web: part of 16M total), but relative scale holds.
Original quote: “[11:21] is the report from Anthropic. I just [11:22] want to very briefly show one thing. The [11:25] scale of Deep Seek's distillation attack [11:27] is just 150,000 exchanges. That is not [11:31] much. Now, Moonshot, the company behind [11:33] Kimmy, had 3.4 4 million and Miniax has [11:37] 13…”
DeepSeek Pro limited now, price drops after 950 supernodes H2 2026 — Verified (95/100)
At 12:30
Nailed the Huawei 950 supernodes plan and compute crunch. DeepSeek's own roadmap 💀✅
Why this score: Spot-on prediction backed by DeepSeek's announcements.
- DeepSeek confirmed Pro runs on Huawei Ascend 950 supernodes, with major price cuts expected H2 2026 after scaling.
- Current throughput is indeed limited by high-end compute constraints per their launch notes.
- Matches web context exactly, no spin needed.
Original quote: “for Pro is very limited. After the 950 super nodes are launched at scale in the second half of this year, the price of Pro is expected to be reduced significantly. So they are very compute constrained.”
DeepSeek V4 nearly as good as Opus 47 and GPT 5.5 — Solid (80/100)
At 13:26
Fair 'nearly SOTA' take — leads open-source, trails closed kings slightly. Accurate flex 😤✅
Why this score: Performance claim holds up per benchmarks.
- DeepSeek-V4-Pro rivals top closed models, leads open-source in math/STEM/coding, closely trails Gemini 3.1 (similar to Opus/GPT tier).
- Described industry-wide as 'nearly state-of-the-art' post-April 24 launch.
- Speaker's 'nearly, not quite' caveat matches real gaps in some frontiers.
Original quote: “Yes, it is nearly state-of-the-art. Nearly, not quite. It is almost as good as the top models Opus 47 and GPT 5.5.”
GPT 5.5 $30/M output tokens, Opus 47 similar price — Verified (100/100)
At 13:57
Prices dead accurate — $30/M for GPT-5.5 output, Opus at $25/M close enough. Chef's kiss 📈✅
Why this score: Pricing facts exactly match current API rates.
- GPT-5.5 base: $30 per million output tokens (confirmed OpenAI April 23, 2026).
- Claude Opus 4.7: $25 per million output (very similar, as stated).
- Speaker rounds up fairly for comparison; DeepSeek V4-Pro is $3.48/M — true fraction.
Original quote: “You see GPT 5.5 is $30 per million output tokens. You see, Opus 47 is similarly priced.”
DeepSeek handles enterprise needs at fraction of US model cost — Opinion (75/100)
At 14:25
Calculus checks out for 90% of biz — why overpay for frontier when 'nearly' wins? Solid CEO bait 🔥
Why this score: Reasonable opinion grounded in real price/performance gaps.
- DeepSeek V4-Pro: $3.48/M output vs GPT-5.5 $30/M (10x cheaper), open-source control perks.
- Performance sufficient for enterprise (not pure frontier research), per benchmarks.
- Strategic take on adoption trends; not universal but logically sound for cost-focused CEOs.
Original quote: “And then you look at DeepSeek and it can accomplish all of your use cases because you're not doing frontier scientific research... And you look at the price and it is literally a fraction.”
US/allied companies will build on Chinese open-source AI — Solid (80/100)
At 14:30
Chinese models hitting 30% global workloads — yeah, the shift is real. US better wake up 😤✅
Why this score: Claim holds up with market data.
- Chinese open-source models like Alibaba's Qwen have 700M+ downloads and 30% of global AI workloads by end-2025 (up from 1% in 2024).
- US policymakers flag this as a security risk with economic displacement potential.
*No exaggeration — trend is documented.*
Original quote: “for a US frontier lab to serve you their model over an open-source Chinese model? And that's where the problem comes in because more and more US and our ally countries enterprise companies are going to think about this and make the decision to build on top of Chinese opensource technology.”
Jensen Huang argued China should use US chips — Solid (75/100)
At 15:05
Jensen's flip-the-script logic lands — fair counterpoint, even if it stings 💀✅
Why this score: Accurate reference to Jensen Huang's public stance.
- NVIDIA CEO Jensen Huang has advocated for China building AI on US chips to maintain supply chain ties.
- Speaker correctly flips it to highlight symmetry in open-source dependency risks.
*Context from recent tech debates confirms.*
Original quote: “Remember Jensen just had the argument that hey China is going to be building their own chips. They're going to be building their own models. They might as well be built on US chips. Well, the same argument is on the flip side with US companies building on top of Chinese open- source models.”
Chinese models pose big US security risk if cut off — Opinion (70/100)
At 15:29
Valid worry — policymakers are screaming about supply chain poisoning already 🔥
Why this score: Plausible opinion backed by expert consensus.
- US concerns include supply chain risks, intelligence collection, and weaker safety in Chinese models.
- Government notes 'industrial-scale' extraction of US AI tech by China.
*Widely discussed in policy circles, not hyperbole.*
Original quote: “That is a big security risk for the United States because if Chinese companies decide to change their architecture or cut us off suddenly, we're in a really bad spot.”
Chinese AI models could dictate global narratives — Opinion (65/100)
At 16:25
AI reflecting creators' politics? Spot on, China vs West divide is real 😬✅
Why this score: Opinion aligns with recent research.
- Feb 2026 study: Chinese models favor state stability/pro-China views; Western ones lean liberal.
- AI shapes info spread, subcultures, narratives per observations.
*Hypothetical but grounded in evidence.*
Original quote: “Imagine we're all built on Chinese models and they're dictating what the models are able to say and what they're not able to say.”
US frontier labs mostly not open-source friendly; Google does small models not matching DeepSeek V4 — Dubious (45/100)
At 16:30
Outdated take — US labs dropping big open models now, not just Google's small fry 💀📉
Why this score: Claim outdated amid recent US open-source releases.
- Google released Gemma 4 family (up to 31B params, Apache 2.0), rivals larger models.
- OpenAI launched gpt-oss-120B, transparent and deployable.
- DeepSeek V4 (preview Apr 24, 2026) is open-source MoE, but US catching up fast per reports.
Original quote: “frontier labs in the US are not open-source friendly for the most part. maybe with the exception of Google, but Google is building small open- source models, not the same level and not the same capability as a DeepSk V4.”
OpenAI/Anthropic models need to get much cheaper fast for US enterprise — Opinion (50/100)
At 16:46
Fair gripe on costs — enterprises hate token bills, but discounts exist 🙄💸
Why this score: Subjective push for faster price drops valid in context.
- Anthropic Claude: $5-25/M tokens, with 50-90% discounts (batch/caching); OpenAI GPT-5.4: $0.20-30/M.
- Still pricier than open-source like DeepSeek; enterprise adoption hinges on cost.
- No hard fact wrong, just opinion on urgency.
Original quote: “we also need to work on efficiency even if we are to maintain closed source and they're being served by OpenAI and anthropic. They need to get much cheaper much more quickly because US enterprise companies need to look at these different models and it needs to make sense costwise.”
DeepSeek doing everything right, Anthropic everything wrong lately — Opinion (50/100)
At 17:04
Bold dunk on Anthropic — 'everything wrong' is spicy rhetoric 🔥😤
Why this score: Hyperbolic opinion, not verifiable fact.
- DeepSeek open-source, cost-efficient (V4 preview Apr 24, 2026); Anthropic faced Feb 2026 fraud accusations.
- Speaker's frustration noted; promotes own video.
- Pure take, no empirical measure.
Original quote: “That's going to enable the entire world to build on top of US artificial intelligence. So, if DeepSeek is doing everything right, Anthropic might be doing everything wrong, at least lately.”
See the full analysis with timestamps →