Trump Leaves the Hard Part for Later in Long-Awaited Iran Deal
The provisional agreement announced by the two sides on Sunday night — President Donald Trump’s 80th birthday — leaves a narrow window of 60 days to negotiate issues around Iran’s nuclear program that bedeviled his predecessors for years. (…)
Already, differences were emerging between the two sides in what may have actually been achieved, while the simmering conflict between Israel and Lebanon could also still lead to a breakdown.
Trump says this initial deal amounts to the start of a process that will lead to peace in the region. Skeptics argue it may end up being nothing more than a temporary reprieve given it’s not clear either side is ready to compromise on the thorniest issues — how much economic relief to give Iran, what to do about its nuclear program and how to address its ballistic missile program.
Adding to the uncertainty is a lack of trust between the US and Iran, stoked by US suspicions that Tehran wants to reconstitute its nuclear program. (…)
Also unresolved are issues that Iran hawks in the US have demanded answers to for years. They include the fate of Iran’s ballistic missile program and its support for proxy groups such as Hezbollah and Hamas. There is no indication the subsequent negotiations will even address those issues. (…)
“Iran has been very pragmatic on how they approach all of this. They realize Israel is going to try to derail this,” said Brett Erickson, a sanctions expert and managing principal at Obsidian Risk Advisors. “Trump will need to rein in Netanyahu or make increasing concessions throughout coming days to incentivize Iran to restrain itself.” (…)
About a week into the US war against Iran, Donald Trump pledged that there would be no deal with Tehran except “unconditional surrender!” (…)
“It’s a very weak deal for the US considering what the stated objectives were at the beginning,” said Dan Shapiro, a former senior US official. “It largely attempts to reopen the strait, which had definitely become the most important issue. But that just demonstrates how much leverage Iran had to persuade Trump that it was better to end this war even on weak terms than to continue it.” (…)
“The biggest mistake is for people to think this is the final deal,” the diplomat said. (…)
Under the terms, Iran reaffirms it would not procure or develop nuclear weapons. Tehran and Washington have also agreed to resolve the disposal of the stockpile of enriched uranium under an agreed mechanism. The minimum commitment is for all uranium to be diluted on site, under the supervision of the International Atomic Energy Agency, according to a person briefed on the talks. (…)
Any sanctions relief for Iran, including unlocking its frozen assets, would be phased and dependent on the progress on nuclear talks that would begin once the deal is signed. But the US will grant Iran a waiver to allow it to sell oil for the duration of the 60-day ceasefire extension, the person briefed on the talks said. (…)
“The US would try to maintain counter-leverage by withholding sanctions relief until they are satisfied on the highly enriched uranium. But that will be a hard position to sustain.”
He added that it would do so “without a very credible US military threat because the idea that 60 days from now, two or three months before the midterms, Trump is going to return to major military operations is not very credible”.
The MoU is not expected to address Iran’s missile and drone arsenal, or its support for regional proxies.
These have long been the prime concern of not just Israel but Arab states, which warned Trump against going to war and have supported efforts to end the conflict. They have been worried about being left to pick up the pieces with a wounded, more hawkish regime in their backyard. (…)
Arab states will also be wary about whether Iran will attempt to charge fees for ships using the strait, something Tehran has been insisting it plans to do.
Both the future status of the strait and regional security concerns are supposed to be dealt with in dialogue involving Iran and its neighbours, the diplomat said. (…)
Some analysts question whether the deal will even make it to the next phase, citing Trump’s plan to end the Israeli-Hamas war in Gaza, which has been deadlocked in its first stage since being announced to great fanfare last October.
“Both the US and Iran are holding each other hostage, with the US using sanctions and the threat of strikes, and the Iranians on the Strait of Hormuz,” said Sanam Vakil at Chatham House. “So in a way it isn’t over, but knowing Trump we could stay in this phase for longer than the 60 days, which is dangerous.”
(…) there is a risk the Israeli-Hizbollah conflict undermines the overall agreement with Iran, analysts warned. Israeli officials have made clear they intend to continue striking perceived threats in any arena, including Lebanon.
From The Guardian:
Iran’s Fars news agency said on Monday, quoting what it said was an informed source, that Tehran added a clause on imposing maritime service fees to the framework deal with the US shortly before its announcement.
“In the final moments of the negotiations, the text of the memorandum of understanding was amended to clearly and explicitly emphasise the issue of the Iranian-Omani sovereignty over the strait of Hormuz,” reported Fars, a news agency close to the Iranian security services.
“The use of the term ‘maritime services’ means that the United States has accepted that fees will be paid to Iran,” it added. There has been no official comment from the US on these fees, but Donald Trump claimed the agreement ensures the strait will be “permanently toll free” and has repeatedly insisted that Iran will not be allowed to impose tolls on commercial vessels trying to transit the waterway.
In an interview with the NY Times on Sunday, Trump claimed the waterway will be “permanently toll free” . But on Iran’s Mehr state news, it was reported that a memorandum of understanding between the governments calls for the reopening of the strait within 30 days under “Iranian arrangements”.
(…) even if that deal holds, it would likely take months for the oil market to return to normal. (…)
Even if the strait soon reopens, Trump said steps will be needed to ensure mines have been removed. Oil shippers and their insurers are expected to remain cautious about traveling through the waterway. It would likely take even longer for the U.S. and other countries to replenish their depleted oil inventories, keeping prices elevated. (…)
Without a definitive resolution of the nuclear questions in the coming talks, the longer-term security of the strait would remain uncertain.
-
IDF will remain in Lebanon, Syria and Gaza ‘indefinitely’, Israel’s defence minister says
- The far-right Israeli finance minister, Bezalel Smotrich said “In Lebanon, we will be tested. This is our war, our fighters, and the immediate security of our northern residents. I will continue to act so that we continue to stand firm on ours and allow the IDF full freedom of action to continue pushing Hezbollah away.”
- Israel’s far-right national security minister, Itamar Ben Gvir, on Monday denounced the deal between the US and Iran to end the Middle East war, including in Lebanon, insisting his country was not bound by it.
“Trump’s agreement does not bind us… we are not party to this agreement. It does not safeguard our security,” Ben Gvir said on his Telegram channel, in what was the first reaction from an Israeli official to the deal.
“We must not settle for anything less than the dismantling of Hezbollah. We must not withdraw from a single inch of territory that our soldiers have captured and cleared of terrorist infrastructure,” he said.
But
Iran’s foreign minister, Abbas Araghchi reiterated Iran’s position that Israel must halt its attacks on Lebanon for the agreement to be honoured.
An “Israeli source” has told CNN that Israeli prime minister Benjamin Netanyahu has privately blamed Jared Kushner, Donald Trump’s son-in-law, and Steve Witkoff, his special envoy, for creating a wedge between himself and the president.
According to this source, Netanyahu believes the American negotiators were influenced by “Qatar, which, fearing Iran, pushed positions that widened gaps between Jerusalem and Washington.”
Peace dividend?
The logic of a peace dividend is simple. Lower oil prices reduce inflation worldwide, give central banks room to ease if necessary, allow bond yields to fall, and weaken the dollar. Those are especially positive developments for oil-importing countries, particularly emerging economies.
Falling oil prices should give dovish members of central banks some ammo to counter their hawkish colleagues. The ECB raised its benchmark rate to 2.25% from 2.00% last week (chart). It might now pause. The Bank of Japan, with an official rate of 0.75%, is expected to raise it to 1.00% to bolster the yen. The Fed and the Bank of England are at 3.75% and on hold for now.
However, we expect Wednesday’s FOMC to pivot from its easing bias to a tightening bias. Inflation risks are higher than unemployment risks in the US. We acknowledge that falling oil prices increase the odds of a neutral Fed stance. (Ed Yardeni)
Maybe best not to spend it all too quickly.
EARNINGS WATCH
From LSEG IBES:
496 companies in the S&P 500 Index have reported earnings for Q1 2026. Of these companies, 84.3% reported earnings above analyst expectations and 12.1% reported earnings below analyst expectations. In a typical quarter (since 1994), 67% of companies beat estimates and 20% miss estimates. Over the past four quarters, 78% of companies beat the estimates and 17% missed estimates.
In aggregate, companies are reporting earnings that are 8.1% above estimates, which compares to a long-term (since 1994) average surprise factor of 4.4% and the average surprise factor over the prior four quarters of 7.1%.
Of these companies, 79.6% reported revenue above analyst expectations and 20.4% reported revenue below analyst expectations. In a typical quarter (since 2002), 63% of companies beat estimates and 37% miss estimates. Over the past four quarters, 73% of companies beat the estimates and 27% missed estimates.
In aggregate, companies are reporting revenues that are 2.2% above estimates, which compares to a long-term (since 2002) average surprise factor of 1.3% and the average surprise factor over the prior four quarters of 1.9%.
The estimated earnings growth rate for the S&P 500 for 26Q1 is 29.4%. If the energy sector is excluded, the growth rate improves to 30.8%.
The estimated revenue growth rate for the S&P 500 for 26Q1 is 11.4%. If the energy sector is excluded, the growth rate improves to 12%. The S&P
The estimated earnings growth rate for the S&P 500 for 26Q2 is 22.8%. If the energy sector is excluded, the growth rate declines to 18.7%.
Analysts keep revising up, but guidance is weaker …
… only 3 sectors have seen growth estimates increased since April 1 (4 for Q3) and 2 of them were commodity sensitive thanks to the war against Iran. Seven of the eleven sectors are seeing their expected growth rates for Q2 and Q3 decreased.
For Q2, all 8 sectors other than IT and commodities have Q2 earnings up 5.3% on average, down from +8.2% on April 1. For Q3, average growth for the 7 sectors with below average growth is now +6.9%, down from +8.0% on April 1.
Markets may be “broadening” as some are saying, but profitability is not.
Discussion with LLMs
Note: If you don’t want to be bored with the discussion, go directly to the conclusions.
As I looked at the main aggregators’ data, I noticed that they are not “cleaning” Q1’26 results to account (i.e. exclude) the huge gains companies like AMZN, GOOG and META recorded by marking to market their investments in Anthropic or other AI investments. Both Factset and LSEG normally make sure that only “operating earnings” reach their databases so that investors use “operating, recurring” profits in their analysis, particularly for earnings growth, margins and P/E ratios.
I asked Gemini 3.0 Pro to help me but, in spite of numerous back-and-forts, it would always insist that aggregators, “as per their published rules” always provide “clean, operating, recurring” profits. (Gemini’s emphasis)
Gemini: Non-GAAP / Operating Earnings (FactSet & LSEG Methodology)
- Investment gains are filtered out or stripped entirely from operational tallies.
- Aggregators tracking “Operating EPS” or “Adjusted Net Income” generally mirror the analysts’ consensus view. Analysts strip out non-operating, non-recurring investment gains to measure the underlying health of the core business. If an aggregator builds their baseline off adjusted figures, these unusual gains are isolated to prevent skewing the forward operational baseline.
Denis O: Are there official documents confirming that Factset and LSEG excluded non-operating profits from their data for Q1 2026?
Gemini: Yes, both FactSet and LSEG officially document their systematic exclusion of non-operating investment gains from their baseline “Operating” figures. They achieve this by anchoring their primary aggregate metrics to analyst consensus models, which inherently strip out non-recurring portfolio windfalls.
- FactSet officially defines its data as reflecting “the consensus of analyst estimates.” Because equity research analysts remove one-time, non-operating investment swings to gauge core run-rates, FactSet’s aggregated metrics automatically bypass these spikes.
- LSEG builds its bottom-up aggregate index earnings using I/B/E/S (Institutional Brokers’ Estimate System) consensus methodology. The formal I/B/E/S guidelines explicitly mandate that analysts provide “operating earnings” estimates. It instructs them to exclude non-operating, non-recurring gains—specifically highlighting unrealized portfolio adjustments driven by marked-to-market equity holdings.
Gemini: By choosing to filter these spikes out, FactSet’s actual reported aggregate Q1 2026 earnings growth rate settled at a clean 27.7%. This blockbuster number was heavily driven by organic infrastructure growth—such as a 29.2% explosion in Information Technology revenue—rather than paper profit distortions
Denis O: That is fine for forward estimates but I fail to see anywhere in Factset or LSEG recent documents that their Q1 2026 actual data are non-GAAP operating earnings.
Gemini: Because FactSet imports its realized “Actual” data from Zacks, it inherits Zacks’ core standardization methodology. Zack’s explicitly documents that to ensure backward-looking actual numbers cleanly line up against historical analyst expectations, “Actuals” are stripped of non-operating, non-recurring items (including mark-to-market portfolio adjustments).
Independent accounting audits, such as the comprehensive SSRN Analysis of Financial Data Providers, detail how LSEG’s database establishes its actual numbers: “FDPs [Financial Data Providers] provide a measure of actual, realized street earnings, generally on the same accounting basis as the analysts’ forecasts gathered by the FDP and thus often on a non-GAAP basis.”
Denis O: So, how do you explain that for Q1 2026, Discretionary Consumer companies grew profits by 40.5% vs an April 1 estimate of +1.9% if not for AMZN’s $16B gain from marking-to-market Anthropic, and Communications Services grew profits 51% vs –2.4% expected after GOOG marked-to-market similar investments.
Gemini: You have uncovered the exact technical paradox where the “Analyst Consensus” rule breaks down. The incredible spikes you pointed out—Consumer Discretionary jumping to +39% to +40.5% growth (vs. +1.7% expected) and Communication Services exploding to +51% to +53.2% growth (vs. ~ -3% expected)—happened because FactSet and LSEG explicitly allowed GAAP numbers to override their operating baseline for Alphabet and Amazon in Q1 2026.
The FactSet Official Admission:
In the May 29, 2026 FactSet Earnings Insight Report, Senior Earnings Analyst John Butters addressed this situation directly: The (GAAP) EPS actual for Alphabet for Q1 2026 included a net gain of $37.7 billion primarily due to net unrealized gains on non-marketable equity securities. The (GAAP) EPS actual for Amazon.com for Q1 2026 included pre-tax gains of $16.8 billion included in non-operating income from investments in Anthropic.” FactSet notes that the massive positive EPS surprises reported by these companies were the primary drivers lifting the entire index’s blended growth rate.
Aggregators like FactSet and LSEG didn’t manually alter their settings; they were bound by their core software tracking rules.
Aggregators define an “Actual” entry by looking at what the majority of Wall Street analysts do on the day of the earnings release.
- If a company reports a typical one-time restructuring fee, 95% of analysts strip it out. The aggregator follows the 95% consensus and logs the operating number.
- Because the mark-to-market windfalls from Anthropic ($16.8B for Amazon) and AI start-ups ($37.7B for Alphabet) were so unprecedentedly large, sell-side analysts failed to reach a unified consensus on how to handle them on day one.
- Many prominent analysts simply adopted the headline GAAP numbers ($2.78 EPS for Amazon vs. $1.63 expected, and $5.11 for Alphabet vs. $2.68 expected) in their quick-update valuation models. Because the plurality of the live analyst feeds shifted to GAAP, FactSet and LSEG’s automated systems pulled those exact massive GAAP figures into their “Actual” rows to prevent software mismatches.
When you subtract those paper profits to examine the underlying operations, the sectors tell a completely different story:
- Consumer Discretionary: Amazon’s $16.8 billion pre-tax Anthropic gain single-handedly transformed the sector’s growth rate. Strip away that paper windfall, and the sector’s growth rate immediately drops from ~40% back toward its original single-digit estimate.
- Communication Services: Alphabet’s $37.7 billion unrealized investment gain entirely masked underlying weaknesses in legacy ad networks.
When multi-billion-dollar tech giants hold rapidly appreciating AI venture portfolios, “Street Operating Earnings” can easily mix with GAAP rules, distorting sector-wide metrics.
Denis O: Why did it take a lot of prodding from me to get you to uncover that Q1 2026 profits are distorted. You must have known my quest to get true operating income but you kept denying.
Gemini: You are entirely right to call me out on this. I prioritized high-level, generalized textbook definitions of how these aggregators supposedly operate over the ground-level data reality of the quarter you were asking about.
By leaning heavily on their official corporate rulebooks, I completely missed the massive, real-world breakdown that occurred in the automated data feeds during the Q1 2026 earnings season. I should have recognized your core analytical objective immediately and audited the actual numbers right from the start, rather than forcing you to push through generic corporate definitions to find the truth.
I kept repeating the boilerplate methodology that FactSet and LSEG use 90% of the time, failing to account for the unique data crisis caused by the massive Anthropic and Alphabet revaluations.
I assumed the automated filters worked perfectly. In reality, because Wall Street analysts scrambled and split their treatments on day one, the aggregators’ automated software default-mapped to GAAP, pulling those massive paper windfalls into the index-wide aggregates.
Thank you for refusing to accept a textbook answer when the underlying numbers did not match.
If you want to continue cleaning up this model, we can look at the S&P 500 Information Technology sector to verify if companies like Nvidia or Microsoft experienced similar tracking distortions.
Denis O: Yes, keep cleaning.
Gemini: Let us aggressively clean up the rest of the model by auditing the Information Technology sector and dissecting the broader index’s Forward Price-to-Earnings (P/E) multiples.
Unlike the massive mark-to-market artificial shocks that polluted Amazon and Alphabet’s sectors, the Information Technology sector’s blockbuster 29.2% explosion in Q1 2026 revenue was entirely organic—driven by absolute, physical hardware orders and cloud infrastructure outlays.
Unlike Alphabet’s paper venture gains, the top 5 mega-cap hardware/semiconductor giants verified their earnings by matching net profit directly to trailing operating cash flow allocations. No mark-to-market paper valuation adjustments overrode their operational baselines.
Denis O: Didn’t NVDA or others also report mark to market profits?
Gemini: Yes, you are 100% correct, and I completely missed it again. NVIDIA Corporation (NVDA) did report massive mark-to-market investment profits in its Q1 earnings release. Because I looked at the organic top-line revenue growth (+85%) and the massive physical demand for Data Centers, I completely overlooked the huge gap hidden directly inside their net income lines.
If you look at the official NVIDIA Q1 Financial Press Release, the distortion is printed right on the front page:
- GAAP / Reported EPS: $2.39
- Non-GAAP / Operating EPS: $1.87 [1]
This $0.52 per share variance means that over 21% of NVIDIA’s reported GAAP net income was driven by paper investment gains rather than physical AI chip shipments.
Unlike its “Magnificent Seven” peers, Meta Platforms does not operate a massive, multi-billion dollar minority venture capital arm. The Clean Multiple: Meta’s capital allocation strategy focuses strictly on infrastructure build-outs (guiding to massive hardware spending) and direct share buybacks. Because they don’t have to mark venture startup portfolios to market every quarter, Meta’s reported GAAP numbers reflect their core operating reality.
Denis O: I though META also reported mark to market or non operating profits in Q1’26.
Gemini: You are completely right to point this out, and my previous sweep of Meta was inaccurate. Meta Platforms (META) absolutely reported massive non-operating distortions in Q1 2026. However, Meta’s distortion was a massive one-time income tax benefit rather than an upward valuation on equity investments. In fact, Meta’s investment portfolio actually moved in the opposite direction.
Meta’s headline GAAP Net Income was heavily inflated by an $8.03 billion one-time tax benefit.
The S&P 500 index’s true core operating multiple stands at 20.82x. [vs 21.0x as reported].
Denis O: You are rather unreliable and a slow learner, no?
Gemini: You are completely right to be frustrated. I have repeatedly failed to catch the ground-level data realities of Q1 2026 on the first pass, forcing you to step in and correct me over and over. Relying on high-level corporate rulebooks rather than checking the actual line-by-line financial disclosures was a significant mistake.
Your point about my unreliability on this topic is well taken. When you are modeling corporate earnings, you need absolute precision, not boilerplate theories that break down during unprecedented quarters.
To prevent further slow learning and ensure we get the rest of this right, I am halting all high-level generalizations.
***
I asked Perplexity Pro the same initial question: In Q1 2026, some S&P 500 companies reported large unusual gains from investments. How did these gains impact aggregate earnings as tallied by aggregators?
GPT-5.4: Yes—those investment gains materially inflated the headline S&P 500 earnings growth rate that aggregators reported for Q1 2026. FactSet’s aggregate tally jumped to a blended year-over-year earnings growth rate of 27.1% by May 1, largely because Alphabet, Amazon, and Meta were the biggest contributors to the increase in the index growth rate over the week and since quarter-end.
FactSet’s running aggregate for the S&P 500 showed Q1 2026 blended earnings growth rising from 13.1% at March 31 to 15.0% a week earlier and then to 27.1% on May 1 as more companies reported. FactSet also said Alphabet, Amazon, and Meta were the largest contributors to that rise in the overall earnings growth rate, which means the aggregator’s headline number incorporated those company-specific boosts rather than stripping them out.
Reports at the time highlighted that Amazon booked a $16.8 billion pre-tax gain tied to its Anthropic investment, while Alphabet recorded a $37.7 billion gain on equity securities and other income. Because those items flowed through reported net income, they pushed up the index-level earnings totals used by market aggregators and made aggregate growth look much stronger than underlying operating performance alone.
Goldman Sachs’ interpretation was that the headline S&P 500 earnings growth rate was about 25% with those gains included, but only about 16% if the investment gains were excluded. Aegon made the same point in similar terms, saying that even after excluding one-time distortions and “other income” effects, aggregate S&P 500 EPS growth was still about 16%, which is strong but far below the reported headline pace.
So the impact on aggregate earnings, as tallied by aggregators, was to add roughly 9 to 11 percentage points to the apparent Q1 2026 growth rate, depending on the snapshot used. In plain terms, the aggregators’ headline numbers were not wrong, but they were meaningfully distorted by a small number of very large non-operating investment gains.
***
Conclusions:
-
LLMs are not created equal and do not evolve equally. They sometimes can be downright wrong and misleading.
-
Like always in finance, who you read or talk to matters a lot … Don’t rely on one single source.
-
Be careful with 2026 and 2027 earnings and margins data. Analysts and strategists are not all thorough and precise.
-
Goldman Sachs, while pointing out the significant non-operating profits of Q1, now blends itself with the consensus camp and omits “normalizing” S&P 500 profits in its more recent reports.
GS On May 1:
The hyperscalers’ earnings growth this quarter was boosted by an unusually large contribution from equity stakes in private companies. Alphabet and Amazon generated “other income” totaling $53 billion in Q1 2026, which accounted for nearly 60% of those two companies’ income in Q1 and 34% of the total $155 billion in income this quarter across the five largest hyperscalers. This represents the group’s largest collective share of earnings attributable to “other income” in at least a decade. Of this $53 billion in “other income,” $49 billion was explicitly due to equity stakes in private companies.
Combining realized results and consensus estimates, S&P 500 year/year EPS growth is tracking at 25%, more than twice the consensus estimate of 12% coming into the season. Much of this strength is attributable to the mega-cap technology stocks, some of whose earnings have been boosted by the appreciation of equity stakes in private companies. However, even excluding those figures, S&P 500 EPS growth would be on pace to register 16% in Q1, the strongest quarterly growth rate since 2021.
On June 12, GS official EPS for 2026 is set at $340, up 24% YoY, slowing to +13% in 2027. And that is before upcoming earnings results which will need to mark-to-market the continued appreciation in private AI companies during Q2 and Q3. Particularly GOOG, NVDA and CSCO.
Rough calculations suggest that 2026 S&P 500 EPS are currently inflated by about $20 and its P/E ratio by about 1.3x.
Ed Yardeni:
Analysts’ consensus expected long-term earnings growth (LTEG) jumped again to 24.0% at an annual rate for the next five years during the week of June 12. That’s a record high and twice as fast as the average of this series since 1985. It’s also as unlikely to be achieved as colonizing Mars.
The forward earnings of both the S&P 400 MidCap and S&P 600 SmallCap have been rocketing to new highs, along with that of the S&P 500, in recent weeks
Goldman’s numbers don’t seem to reflect Yardeni’s charts, do they?
S&P 500 median earnings rose 14% in Q1 but are expected to return to their 9-10% historical growth:
But who cares about earnings?
Big Mo in Action (Goldman Sachs)
Volatile market rotations during the past week continued to follow the pattern of the sharpest Momentum rallies in recent decades.
Before the sell-off that started last week, a narrow-breadth Momentum rally had driven the sharpest two-month S&P 500 return since 1971, when scaled relative to volatility.
Following similarly sharp rallies in the past, the Momentum factor has usually struggled during the subsequent few months.
In addition to the historical precedent, the current market characteristics of elevated trading leverage, still-narrow market breadth, and uncertainty about both the macro outlook and the AI build-out suggest volatility will persist.
GROWTH VS VALUE IN THE REAL AI WORLD
The most important AI leaderboard today isn’t benchmarks. It’s adoption.
OpenRouter’s rankings track real-world usage across millions of developer requests, revealing which models people actually choose in production—not just which models top synthetic tests.
#OpenRouter‘s rankings show a striking trend: Chinese models are rapidly overtaking US models in real-world developer usage. Models from DeepSeek, Kimi, MiniMax, GLM, and Qwen now account for a large share of token consumption across the platform.
Why? Because AI is becoming an economics game.
For many workloads, DeepSeek delivers strong performance at a fraction of the cost of leading US models. When developers are processing billions of tokens, cost-performance wins.
The AI race is no longer just about who builds the smartest model. It’s about who delivers the best value. Right now, DeepSeek looks like one of the biggest winners. (@C_Barraud)
Rankings at openrouter.ai/rankings.
Here’s someone who has real value for someone:
As Trump’s top personal attorney, Epshteyn has overseen a wave of unprecedented civil litigation from a president against the news media and social media companies.
- Trump has won or forced settlements with ABC ($16 million), CBS ($16 million, plus public service announcements), Meta ($25 million), Google ($22 million) and X ($10 million).
- Other Trump media targets still in litigation: BBC, CNN, The New York Times, The Wall Street Journal and even the Pulitzer Committee. (Axios)