Grok 4.6 tops CursorBench at $2.81 per task, far below rivals

As seen on the 24/7 Wall St. homepage on August 21, 2026.

Elon Musk @elonmusk Quote post

The number that stings rivals is $2.81 per task versus $17.32 for the model Musk says Grok just edged out. Vendor benchmark claims deserve scrutiny, but coding-model pricing is where AI margins get decided.

Grok 4.6 on extra high thinking mode now achieves #1 score on CursorBench! https://t.co/MZDjMTtqK2 [Quoted @XFreeze]: Grok 4.6 just took the #1 spot on CursorBench 3.2.....and the efficiency is insane Here's the cost comparison: • Grok 4.6 Extra High — 70.8% | $2.81/task • Fable 5 Max — 70.5% | $17.32/task • Opus 5 Max — 70.0% | $8.23/task • GPT-5.6 Sol Max — 67.2% | $5.69/task Grok https://t.co/gZXRhuXlka
  • Replies343
  • Reposts192
  • Likes1.4K
Continue ReadingShow less

Grok 4.6 in extra high thinking mode took the top spot on CursorBench 3.2 with a score of 70.8%, ahead of Fable 5 Max at 70.5%.

Grok 4.6 Extra High priced out at $2.81 per task against Fable 5 Max's $17.32, so developers can run roughly six Grok tasks for the price of one from the closest benchmark rival.

CursorBench is a coding-focused evaluation, so for companies deploying AI at scale in software development workflows, per-task cost directly shapes margins.

Elon Musk amplified the result on X on August 21, quoting an account that had tabulated the benchmark standings. Vendor-amplified benchmark results warrant independent scrutiny, but the figures give xAI a concrete efficiency story for enterprise customers weighing coding model contracts.

Sources

Related Market Updates

Chamath Palihapitiya @chamath

Chamath is framing Anthropic's safety pitch as a moat play, landing just as the company lines up a reported IPO and a potential Nvidia investment.

“We must slow the pace at which we improve the capabilities of AI models.” Dario makes the case to stop open source and concentrate enormous technological and economic power with Anthropic. https://t.co/ILhs01iQ8u
  • Replies9
  • Reposts2
  • Likes19
Read more
Elon Musk @elonmusk Repost

Musk endorsing the idea that Full Self-Driving now drives North American vehicle demand moves the Tesla case from units to software attach rates. Watch for an FSD take rate number on the next earnings report.

RT @SawyerMerritt: I think people are significantly underestimating just how much FSD has become a major driver of Tesla sales in North Ame…
  • Replies0
  • Reposts440
  • Likes0
Read more
Artificial Analysis @artificialanlys

Ant Group is giving away weights that hit the efficiency frontier with just 5.5 billion active parameters, which is the kind of cost curve that squeezes pricing for closed US model vendors.

Ling-3.0-flash-VL, Ant Group’s new flash tier open weights model, scores 25 on the Artificial Analysis Intelligence Index. At 124B total parameters 5.5B active parameters, it sits on the Intelligence vs. Active Parameter Pareto Frontier @AntLingAGI has released https://t.co/XjfdrMhf0H
  • Replies1
  • Reposts0
  • Likes25
Read more
Artificial Analysis @artificialanlys Reply

OpenAI held image generation at $30 per million output tokens while raising quality, squeezing rivals who now have to match better output at an unchanged price.

GPT Images 2.5 are a material improvement in quality at the same price as GPT Image 2 OpenAI prices both new models at the same token rates as GPT Image 2: $30 per million image output tokens. This makes it the clear choice for the highest quality images, though the next model https://t.co/GvxSEkjj6p
  • Replies1
  • Reposts0
  • Likes2
Read more
Artificial Analysis @artificialanlys

OpenAI just halved image generation time at the same price, resetting the cost-per-image bar rivals like Google and Midjourney have to match.

OpenAI’s GPT Images 2.5 takes the top two spots on the Artificial Analysis Image Arena, generating higher quality images in less than half the time, while keeping the same price as GPT Image 2 Nearly five months after GPT Image 2’s release, OpenAI has released GPT Image 2.5 https://t.co/B8BItbLs7H
  • Replies2
  • Reposts0
  • Likes16
Read more
PODCAST Dwarkesh Podcast 1 hr 37 min

AI researchers debate how close we are to recursive self-improvement

  • Panel steelmans failure: Moravec's paradox, no continual learning, a persistent sim-to-real gap blocking generalization
  • Schulman on distillation from Claude creating an open-weight monoculture, and Chinese labs closing the gap via router proxy data
  • Cursor already runs online RL off user acceptance signals every five hours
  • Rapid-fire: 1-3 years to drop-in AI remote workers, 2 years to 10x researcher uplift, 3-10 years to beating top experts

Frontier researchers, including Thinking Machines chief scientist John Schulman, put drop-in AI remote workers at one to three years away, which is the timeline underwriting the entire AI capex trade. Their bear case is technical: RL generalizes task length rather than reasoning across domains, even with task completion time doubling every three months per Edgebench.

Read more
Elon Musk @elonmusk Repost

Tesla is taking the Semi into Europe with a claimed 550km of range at a full 40 tonnes, opening a second front for a truck program that has barely started volume deliveries at home.

RT @teslaeurope: Semi is coming to Europe – Up to 550km of range at 40T (fully loaded) – 1 kWh/km energy consumption – Megacharger at up t…
  • Replies0
  • Reposts394
  • Likes0
Read more
Artificial Analysis @artificialanlys

A model roughly a third the size beating DeepSeek's own 1.6T flagship at about a quarter of the cost per token undercuts the compute-scaling logic funding the AI infrastructure trade.

DeepSeek V4.1 Flash overtakes DeepSeek V4 Pro 0813 as DeepSeek’s new flagship model with a score of 40 on Artificial Analysis Intelligence Index. At just 552B parameters, it outperforms the Pro (1.6T) model while costing ~4x less per token, placing it just short of the https://t.co/ne5qNtijJ5
  • Replies12
  • Reposts11
  • Likes114
Read more
Jim Cramer @jimcramer Quote post

Gasoline at $4.28 a gallon pulls 0.5% out of disposable income, money that discretionary retailers and restaurants were counting on this quarter.

$118 shock https://t.co/QyqjlGaD7x [Quoted @KevRGordon]: .@RenMacLLC: “Retail gasoline prices are $4.28 per gallon on average…$1.08 per gallon higher than the same day last year. Doing the math, this works out to a $118 billion shock to disposable income, or 0.5%.” https://t.co/XEQXAjzw1G
  • Replies26
  • Reposts4
  • Likes57
Read more
Jim Cramer @jimcramer

Cramer is pointing at the AI risk nobody prices in: insurers backing away from copyright and liability exposure, leaving the leading labs to absorb litigation costs directly. That bill lands on the private valuations investors are chasing.

Can OpenAI and Anthropic get insurance?? Long before they kill us, the plaintiffs' lawyers will kill them without it
  • Replies19
  • Reposts2
  • Likes48
Read more
HOUSE TRADE DISCLOSURE Gilbert Ray Cisneros, Jr.
  • AMD Advanced Micro Devices, Inc. (AMD)Partial sell$1K – $15K
  • AGIO 150 Main Street Trust > Bank of America Agios Pharmaceuticals, Inc. - Common StoBuy$1K – $15K
  • APH 150 Main Street Trust > Bank of America Amphenol Corporation Common Stock (APH)Buy$1K – $15K
  • APG 150 Main Street Trust > Bank of America APi Group Corporation Common Stock (APG)Buy$1K – $15K
  • APP 150 Main Street Trust > Bank of America Applovin Corporation - Class A Common StPartial sell$1K – $15K
  • AORT 150 Main Street Trust > Bank of America Artivion, Inc. Common Stock (AORT)Buy$1K – $15K
  • ATRC AtriCure, Inc. - Common Stock (ATRC)Buy$1K – $15K
  • BSX 150 Main Street Trust > Bank of America Boston Scientific Corporation Common StoBuy$1K – $15K
+ 42 more trades in this filing
Disclosed 37 days after the earliest trade

Fifty-eight transactions in one filing amount to a full portfolio rebuild: bigger-ticket buys in Microsoft, Eaton, Quanta Services and Globus Medical, funded partly by repeated Lamar Advertising sales and trims in AMD, Nvidia and Meta. The rotation leans toward power, infrastructure and medical devices while chip and ad-tech exposure comes down. Disclosed 37 days after the earliest trades.

Read more
Read more