K Koda Intelligence
Subscribe
Cover image for the 02 October 2026 Signal: By Google's count, Gemini 4 Argon beats Claude Opus 5.5 on the Vals Index while matching GPT-6.1 Sol's discounted price.
THE SIGNAL · 02 OCT 2026 · 10 MIN READ

Google launches Gemini 4 Argon at $2 per million input tokens, matching GPT-6.1 Sol

Google priced Gemini 4 Argon at $21 per million input tokens and $102 per million output tokens. It claims a 68.9%3 Vals Index score, ahead of Claude Opus 5.5 and GPT-6 Astra, so today's real question is what a call actually costs.

KI Koda Editorial · Updated 00:54 GMT · 24 sources
AIMODELSPRICING
44 of 50 figures cited16 receipts in the register
  1. $21CNBC
  2. $102CNBC
  3. 68.9%3CNBC
The register
Lead Story

The day's defining move.

Model Release · CNBC
Lead storyModel Release $21 CNBC · 01 OCT 2026
Model Release·CNBC·01 OCT 2026

Gemini 4 arrives amid Wall Street's agent shift

Google priced Gemini 4 Argon, announced September 30, at an introductory $21 per million input tokens and $102 per million output tokens, matching OpenAI's discounted GPT-6.1 Sol. Google says Argon scored 68.9%3 on the Vals Index, ahead of Anthropic's Claude Opus 5.5 at approximately 67%4 and OpenAI's GPT-6 Astra at 63.1%5.

Why it mattersCheck the $102 output rate against Argon's 1 million output-token ceiling: one maximal response costs about $102, so long-generation jobs need per-call budget caps.

Continue reading
AI

Google priced Gemini 4 Argon at an introductory $21 per million input tokens and $102 per million output tokens, and says it scored 68.9%3 on the Vals Index, ahead of Claude Opus 5.5 at 67%4 and GPT-6 Astra at 63.1%5.

World

Roughly 9,0006 US troops and a third aircraft carrier are heading to the Middle East as Trump weighs new strikes, after he first promised an Iran campaign lasting weeks.

Markets

The market mood gauge reads Greed, signaling investors are leaning toward risk-taking.

Wild Card

Washington is moving closer to frontier AI launches: Google says Gemini 4 Argon is part of the Trump administration's pre-release access program, and OpenAI's Greg Brockman signed the White House-backed Joint Commitment on Frontier Responsibilities alongside Google, Anthropic, Meta and Nvidia executives.

KODA PROFrom the desk

The Operator Tier is coming.

A weekly operator deep dive, the full Dojo Pro prompt packs, and the complete prompt database. Founding members lock the launch price forever.

Join the founding list →
Listen & Watch

Daily broadcasts.

Podcast · Video · Infographic
YouTube · Short

Today's Signal Short

Visual

Intelligence map

Markets

6 levels at the close.

Market TerminalLAST CLOSE 01 OCT 2026 · READ 2026-10-02 · Yahoo Finance
S&P 500 7,666.4510 +0.19%10 6D, window short 7,651.54 low / 7,743.41 high
Nasdaq 26,871.6011 +0.04%11 6D, window short 26,797.54 low / 27,068.72 high
Bitcoin $84,821.1612 +1.51%12 6D, window short $83,502.61 low / $84,817.76 high
Ethereum $2,704.4313 +0.77%13 6D, window short $2,676.72 low / $2,704.43 high
Crude Oil (WTI) $92.9214 +2.76%14 6D, window short $89.38 low / $94.61 high
Crypto Fear & Greed 72 / 10015 Greed 0 to 100 index 25 fear / 50 neutral / 75 greed · alternative.me
Today's Focus

What else the day turns on.

01

Kalshi's Proposed $1 Billion7 Round

Kalshi's proposed $1 billion7 round, with Tiger Global and Dragoneer in talks to invest, could value the prediction-market exchange at about $40 billion, up from $22 billion in May. The financing has not closed. The company is pushing past event contracts toward a broader trading platform, which puts it closer to established exchanges.

Source: tldr.tech
02

Brockman Signs Frontier Commitment

OpenAI President Greg Brockman signed the White House-backed Joint Commitment on Frontier Responsibilities on Oct. 1, alongside executives from Google, Anthropic, Meta and Nvidia. Signatories pledged stronger controls on cybersecurity, biosecurity and chemical risks, plus independent audits.

03

Kled AI's 72-Hour Datasets

Kled AI's 500,0008 volunteer contributors act out tasks to produce custom training datasets. The company's selling point is speed: delivery within 72 hours.

Source: archive.superhuman.ai
Deep Dive

Today's editorial.

All editorials
Deep Dive cover image: Your $2 Tokens Will Cost $4
Deep Dive6 min read

Your $21 Tokens Will Cost $4

Read full analysis
The Wire

8 stories from 6 desks.

Lead carried above
Policy·CNBC·30 SEP 2026

Google unveils Gemini 4 Argon AI model

Google says Gemini 4 Argon is part of the Trump administration's pre-release access program for AI models.

Why it mattersGovernment pre-release review adds a gate before general availability, so Google's still-unannounced public launch date depends partly on federal testing.

Read
Hardware·CNBC·01 OCT 2026

Alphabet Sends In-House AI Chips Into Orbit

Alphabet is preparing to launch its proprietary AI chips into orbit to test whether computing hardware can run in space. The payload will fly on a SpaceX Falcon 9.

Why it mattersTreat space-based compute as a research line item until Alphabet publishes on-orbit performance data, since no benchmark or uptime result exists yet.

Read
Hardware·TechCrunch·30 SEP 2026

DoorDash Air Unveils Its Own Delivery Drone

On September 30, DoorDash Air unveiled a purpose-built autonomous aircraft, bundling the drone, ground operations, marketplace and automated flying into one delivery service for local businesses.

Why it mattersLocal restaurants and retailers on DoorDash should ask for per-delivery drone fees before signing up, since the announcement named no pricing or service areas.

Read
Also on the wire
  1. AnthropicBarclays Scales Anthropic's Claude Across Operations
  2. CNBCUS Order Renames AI as 'Super Intelligence'
  3. The VergeProtesters in Robot Costumes Greet OpenAI DevDay
  4. OpenAIAlbertsons Rebuilds Retail Operations With OpenAI
  5. NVIDIACoreWeave targets agentic AI with Vera Rubin
The Globe

3 stories off the AI beat.

Conflict·WSJ·01 OCT 2026

US Sends 9,0006 Troops, Third Carrier Mideast

Trump first promised an Iran campaign lasting weeks. Now roughly 9,0006 troops and a third aircraft carrier are heading to the Middle East as he weighs new strikes.

Why it mattersTrump says escalation after the November midterms remains an option, so fuel and freight budgets should carry a Gulf risk premium into early 2027.

Read
Also on the globe
  1. BloombergPutin Ties Kaliningrad Defense to Nuclear Arsenal
  2. France24US tells European allies to lower diesel prices
The Lab

Tool of the day, researched.

All Lab reports
Anything home page, captured for the 2 October 2026 Lab report Tool of the day 6.316/ 10
Coding

Anything

Anything will spin up a good-looking app from a paragraph; whether that app still remembers your data after a sign-out is the part worth checking.

Every Anything project ships with its own Postgres database and Stripe payments wired in, which puts it ahead of builders that stop at a pretty front end. User reports split on reliability, so treat the first build as an experiment. Non-technical founders validating an idea should adopt it, starting on Free and moving to Pro only once a test build keeps its data.

Capability 7.0
Ease 7.5
Value 5.0
Momentum 5.5

Built two apps so far and found the native database wonderful.Paraphrased · Product Hunt

Read the Lab report Try it
Also on the radar
Productivity

NeuralAgent: a voice-driven assistant that operates your whole PC

NeuralAgent reads your screen and controls your computer directly. It opens apps, clicks buttons, manages files, and carries out tasks across your system from voice commands. Start with a repetitive file-cleanup or app-switching chore on a machine without sensitive logins, and watch each step before handing it anything that touches money or credentials. It is free for limited access, and extra features are pay-as-you-go.

Try it
Creativity

Loova AI: one studio for UGC videos, avatar ads, and product shots

Loova Ads Studio generates UGC-style videos, product commercials, avatar videos, and static creatives in one place. Feed it a single product and build three ad variants, one UGC clip, one avatar spot, and one static image. Run them as a small test so your audience picks the winning format before you pay for a full shoot. It is listed as free.

Try it
Mindset

Brev.io: an AI coworker that tracks whether goals are actually moving

Brev connects goals, meetings, metrics, and decisions across the tools your team already uses, including Slack and Teams, and shows leaders where execution stands in real time. Load one quarterly goal and its owners first, then check after two weeks whether the status matches what your standups report. It is listed as free.

Try it
Creativity

neolocus.ai: generate and edit interior and architectural renders at scale

neolocus.ai is a rendering playground for interior designers and architects that produces and edits renders in batches. Use it early in a client engagement to show five material or layout directions from one base image, then reserve full manual renders for the option the client chooses. Confirm the pricing on the listing before you build it into a project quote.

Try it
The Arena

3 labs moved this week.

  1. OpenAISigns White House safety accord as FTC reportedly opens AI-safety probe fortune.com · 01 OCT 202601
  2. AnthropicBarclays scales Claude across operations as Amodei signs frontier safety pledge anthropic.com · 01 OCT 202602
  3. Meta AIZuckerberg signs voluntary White House pledge on frontier model safeguards fortune.com · 01 OCT 202603
Builder Radar

One paper, read at the source.

Primary sources, pulled by index
Paper of the day

AgentHop: A Diagnostic Benchmark for Agentic Multi-Hop Scientific Question Answering

Chanhee Park, Jeongho Yoon, Sungbin Han et al. · arXiv 2026-09-28

Agentic tasks require a large language model to interact with the world, navigating information and gathering evidence across multiple steps with restricted resources.

Conversion cells are shaded only where a paired same-item test separates the tier leader from the runner-up, as reported in Appendix E.1 ‣ Appendix E Diagnostic Detail and Robustness Checks ‣ AgentHop: A Diagnostic Benchmark for Agentic Multi-Hop Scientific...
Read the paper

The paper comes from the Firecrawl Research Index: arXiv, indexed in the last 7 days and submitted this month or last, picked for today's lead. Quotes are verbatim.

THE DOJO · BUILD TODAY

Put a hard ceiling on every Argon call before you ship

  1. 01

    Cap output tokens per request. Argon allows up to 1 million output tokens, so one maximal response costs about $10 at the output rate. Set a max-output limit on every long-generation job.

  2. 02

    Trial one Anything build end to end. Create a small project and check that its Postgres tables and Stripe test payments work before you commit real users to it.

  3. 03

    Scope a custom dataset with Kled AI. Write one narrow task spec and test whether a 72-hour turnaround from its volunteer contributors fits your fine-tuning schedule.

The Bottom Line

Argon's sticker price is the floor. Your per-call cap sets the ceiling.

Google set Gemini 4 Argon at $21 per million input tokens and $102 per million output tokens. That launch is exactly the kind of introductory rate the deep dive warns can double in practice once long outputs pile up. With a 1 million output-token ceiling, one maximal Argon response costs about $102. Set per-call budget caps and measure your blended cost per task before you move workloads over.

Source: cnbc.com
Receipts

Every number, and where it came from.

44 of 50 figures cited, 16 receipts: 4 verified, 9 reported, 3 estimated
  1. 01$2...Gemini 4 Argon, announced September 30, at an introductory $2 per million input tokens and $10 per million output...CNBC01 OCT 202603B search corroborationV
  2. 02$10...September 30, at an introductory $2 per million input tokens and $10 per million output tokens, matching OpenAI's...CNBC01 OCT 202603B search corroborationV
  3. 0368.9%Google says Argon scored 68.9% on the Vals Index, ahead of Anthropic's Claude Opus 5.5 at approximately 67% and...CNBC01 OCT 202603B search corroborationV
  4. 0467%...Index, ahead of Anthropic's Claude Opus 5.5 at approximately 67% and OpenAI's GPT-6 Astra at 63.1%.CNBC01 OCT 202603B search corroborationE
  5. 0563.1%...Opus 5.5 at approximately 67% and OpenAI's GPT-6 Astra at 63.1%.CNBC01 OCT 202603B search corroborationE
  6. 069,000US Sends 9,000 Troops, Third Carrier Mideast.WSJ01 OCT 202603B search corroborationV
  7. 07$1 billionTiger Global, Dragoneer in talks to invest in Kalshi's new $1 billion fundraiseTLDR01 OCT 2026as publishedR
  8. 08500,000Kled AI's 500,000 volunteer contributors act out tasks to deliver custom training datasets within 72 hoursKled AI, via Superhuman01 OCT 2026as publishedR
  9. 09$100...four tiers, from a free plan with stricter Grok limits to a $100 per month Ultra tier with the Grok Bot agentTLDR01 OCT 2026as publishedR
  10. 107,666.45S&P 500 close, +0.19% on the sessionYahoo Finance01 OCT 2026yfinance closeR
  11. 1126,871.60Nasdaq close, +0.04% on the sessionYahoo Finance01 OCT 2026yfinance closeR
  12. 12$84,821.16Bitcoin close, +1.51% on the sessionYahoo Finance02 OCT 2026yfinance closeR
  13. 13$2,704.43Ethereum close, +0.77% on the sessionYahoo Finance02 OCT 2026yfinance closeR
  14. 14$92.92Crude Oil (WTI) close, +2.76% on the sessionYahoo Finance01 OCT 2026yfinance closeR
  15. 1572 / 100Crypto Fear & Greed close, Greed on the sessionalternative.me02 OCT 2026as publishedR
  16. 166.3 / 10Anything, Koda score across four dimensions04R dossier02 OCT 202604R dossier scoringE
V
Verified: the stat gate corroborated this figure against an independent search before publication.
R
Reported: the figure is carried as published by the linked source and was not independently corroborated.
E
Estimated: the figure is approximate or hedged, either in the source or by the stat gate.

Method names the computation path behind the row, from a closed vocabulary: as published, yfinance close, 03B search corroboration, 04R dossier scoring, not recorded.

A figure is money, a percentage, a spelled magnitude, a multiple, a rate carrying its unit, a score over its denominator, a thousands separated number or an index close. Bare integers and years are not counted, so a phrase like "over 100 integrations" carries no mark. A chart's own scale line is the axis of the level above it and is counted once, with that level.

Get the morning Signal

189 editions so far, one a day. Unsubscribe anytime.

Forward this to one operator you work with. Your referral link is in every email; milestones at koda.community/refer.