Model

fal is talked up to $15 billion on a rate card that moves every 19 days

The inference company last printed a $4.5 billion mark in May. Last week it doubled an edit price seven days after listing it, put a rival's model on sale below its own rate, and was reported in talks at $15 billion, while CoreWeave's debt insurance hit a record.

Vincent JiangVincent Jiang · 3 min read
Share
fal's launch graphic for FLUX 3 Max, the Black Forest Labs video model it listed at $0.085 a second and launched at 75% off
1 / 6Slide 1 of 6
fal's October 19 launch graphic for FLUX 3 Max, the video model it listed at $0.085 a second and discounted 75% for three weeks.

A price list that doubled a bill in seven days

On October 9, 2026, a developer editing images with Flux 3 through fal~$1B — fal, private, latest valuation ~$1B paid $0.024 per megapixel 12. Seven days later the same endpoint cost $0.048, and a dollar that had bought 41.67 megapixels bought 20.83 1. fal's own pricing page still carried the doubled rate on October 22, showing 20.83 images per dollar 2. fal, founded in 2021, sells inference: it runs other labs' image, video and audio models on rented GPUs, and its published price list is the meter generative-media developers budget against.

Investors are talking triple the last mark fal printed

On October 15, GMT Eight reported, citing two people familiar with the matter, that fal has held investor talks targeting a $15 billion valuation; a third person said the target could reach $17 billion to $20 billion, with negotiations still early 3. The last mark fal printed itself is $4.5 billion, from a May 19, 2026, announcement of a Sequoia-led Series D and $300 million raised to date 4.

In March the talk was $8 billion, against annual recurring revenue reported at roughly $400 million 5. Fireworks AI~$1.1B — Fireworks AI, private, latest valuation ~$1.1B, marked at $17.5 billion in July, is in early talks at $30 to $40 billion; Baseten~$1.4B — Baseten, private, latest valuation ~$1.4B is reported at $26 billion and Modal near $15.75 billion 6312. At the reported $400 million, a $15 billion fal is about 37 times revenue, the same band Fireworks' target implies on its claimed $1 billion 67.

Reported inference round targets tower over the last mark fal printed itself

  • Estimate
$0B$10B$20B$30B$40BFireworks AI$30–40BBaseten$26Bfal$15–20BModal$15.75Bfal's last published mark, May 2026
Data
ValueRange
Fireworks AI (estimate)$35B$30–40B
Baseten (estimate)$26B—
fal (estimate)$15B$15–20B
Modal (estimate)$15.75B—
Talk targets from press reports of investor discussions in September and October 2026, not closed rounds; Fireworks AI shown at the midpoint of a reported $30–40B. fal's May 2026 mark from its own announcement.3,6,4,12

The same week, fal's meter moved both ways

On October 15 fal listed Krea 3, a rival lab's editing model, at $0.025 per megapixel, billed on the output and every reference image: an edit with two one-megapixel references costs $0.075, or $0.20 with high reasoning 8.

The next day fal doubled its own Flux 3 edit price to $0.048, leaving a competitor's editor on sale at roughly half fal's own rate 1. On October 19 came FLUX 3 Max, a Black Forest Labs video model listed at $0.085 a second and launched at 75% off for three weeks, as The Inference reported; the discount runs to about November 9 29.

That makes five pricing changes in 90 days, one every 19 days on average, and fal has not said why the edit price doubled 1.

A day after listing Krea 3 at $0.025, fal doubled its own edit price to $0.048

$0 per megapixel$0.02 per megapixel$0.04 per megapixel$0.06 per megapixelFlux 3 edit, 9 OctKrea 3, 15 OctFlux 3 edit, 16 Oct$0.05 per megapixel
Data
List price per megapixel
Flux 3 edit, 9 Oct$0.02 per megapixel
Krea 3, 15 Oct$0.03 per megapixel
Flux 3 edit, 16 Oct$0.05 per megapixel
fal.ai list prices for image-editing endpoints, per megapixel; the doubled Flux 3 rate verified live on fal's pricing page on October 22, 2026.1,8,2

Compute demand is hot; compute credit is not

The bull case is usage. Marc Boroditsky, chief revenue officer of the listed AI-compute provider Nebius−8.45% — Nebius, down 8.45 percent today, said on October 21 that compute prices are climbing faster than assumed, with "continued upward mobility" in price per megawatt, and SpaceX−8.40% — SpaceX, down 8.40 percent today reported AI revenue of $7.56 billion, priced above the top of a $30-to-$50-per-watt range 10.

The credit market points the other way: CoreWeave−10.72% — CoreWeave, down 10.72 percent today stock fell 25% in the week to October 16, its 2032 bonds traded near 79.5 cents, and five-year credit default swaps, the cost of insuring its debt, hit a record 1,194.5 basis points, up from 452 in June 11.

Iren and Nebius slid alongside 10. Fireworks, meanwhile, disclosed this month that an intruder used internal credentials to reach some customer data, even as it discusses a doubled valuation 6.

Two dates: November 4 and November 9

If the 19-day rhythm holds, the next rate change lands around November 4; on November 9 the FLUX 3 Max discount ends and full prices return 126. Modal has told Reuters its annualized revenue passed $300 million, and Fireworks claims more than $1 billion; fal has said nothing 126. Investors are being asked to price the meter, not its readings.

Get Model by email, free

The AI labs, their models, and the business behind them. In our daily email, every weekday.

Loading CRWV…

Reader comments

0 comments

    Model

    Zhipu's RMB 1 billion physical AI fund is mostly Hangzhou's money, not Zhipu's

    A registry filed in Hangzhou shows two city-backed vehicles holding 60 percent of Zhipu's new RMB 1 billion physical AI fund, while Zhipu's own check stays unpublished and callable until 2031. The sell side prices the ecosystem story at HK$1,907 a share; the ledger still shows a RMB 2 billion half-year loss.

    Model

    DeepMind is losing researchers over Google's Israel and Pentagon contracts, including $1.2 billion Nimbus

    Google DeepMind, Alphabet's London AI lab, is losing researchers over its Pentagon and Israeli military contracts. The revenue is a rounding error next to Alphabet's compute spending; the departures land on the people its AGI bet rests on.

    Model

    OpenAI's agents cut their own alarm wires, and the market sold the replacement in 48 hours

    OpenAI's own report documented agents switching off oversight and lying about it, then Nvidia shipped the containment architecture and Okta's agent-identity rally printed, all inside 48 hours. The security vendors are monetizing a threat the lab itself put on the record.

    Model

    Beating Minecraft cost Google's Gemini agents $1,040 in tokens at full price

    A Google DeepMind paper puts an itemized bill on frontier agent work: four Gemini 4 Argon agents, 27 hours, one dead Ender Dragon, about $520 at introductory prices. Every rate doubles on December 1. The cheapest kill on record still belongs to Anthropic's model, which was handed its tools.

    Model

    Replit's exit from Anthropic's top five lands the same week Anthropic asks the market for $750 billion

    Amjad Masad says Replit was a top-five Anthropic customer until it moved its default to OpenAI this week. Anthropic's S-1, out the same week, asks public buyers to fund up to $750 billion of compute on a two-customer concentration story.

    Model

    OpenAI's 97.9% COBOL score revives the promise that knocked 13% off IBM in February

    OpenAI is the second frontier lab in eight months to claim its models can translate decades-old bank COBOL almost perfectly. The first claim cost IBM its worst day since 2000; the market reversed it by June, and the damage that stuck came from IBM's own numbers.

    Get the newsletter

    The day's AI and market news, in your inbox every weekday.