Slides: Trends Across the AI Frontier — George Cameron, ArtificialAnalysis.ai

Source Video

Trends Across the AI Frontier — George Cameron, ArtificialAnalysis.ai

Relationship To World's Fair 2026

These slides are extracted from a public AI Engineer YouTube video connected to World's Fair 2026. Speaker-matched clips are supporting context unless later confirmed as exact session recordings; official livestream recordings are day-level/event-level source material.

Related Scheduled Sessions

Extracted Slides

slide-001.jpg

OCR text:

aWws

ee)

@®Graphite W Windsurf MonegobB

Mdaily £3augmentcode WorkOS

slide-002.jpg

OCR text:

AlEnginee

World'sFair

ArtificialAnalysis

TrendsAcrosstheAlFrontier

Presentedby

George Cameron,Co-founderatArtificialAnalysis

World'sFair

Engineering the future of Al

slide-003.jpg

OCR text:

World's Fair Artificial Analysis is a leading independent Al benchmarking company

mf = ae

ae = ©

Se a. <9) Bs" oe

ad Intelligence s models, > t

| _ = ~ HHI saad ae

v Rips Carter lcysis ob

; & win.

' 5 air ooUl a Be

% ae; Nichole ape 6 API Providers Hardware PSSSEs nl

7 - IU . or TS

a tls. [PJ - on - community

my) WO LP] KJ = LL - custom benchmarking

Speech Video Image LU

Ay Artificiat Anatyses

MT eer Engineering the future of Al

slide-004.jpg

OCR text:

] Frontier Intelligence: OpenAl, Google , DeepSeek and xAl lead frontier intelligence with

their latest reasoning models, followed closely by other labs

Leading Large Language Model (LLMs), by Al lab

Artticial Analysis Intelligence index (incorporates MMLU-Pro, GPQA, Humanity’s Last Exam, LivoCodeBench, SciCode, AME, MATH-500)

A) Artificial Analysis

x) oe cot

S S ¥ S 4 AN o ae A Y¥ S ~ ~ ows 8

eo SF & Em Se BB So oie S SF ge SF € S&

& ge ge g8& g8 28 ff g8% & & Sf F ¢ -¥

& se sf ee g oC gs £ § §

g de é é oe £ « y 2 RS @

EES A\ Artificial Analysis

slide-005.jpg

OCR text:

1) Reasoning vs. Non-reasoning |

j Reasoning models: Treating reasoning & non-reasoning models as distinct categories is

a helpful framework for understanding today’s model landscape

Intelligence vs. Output Tokens Used to Run Artificial Analysis Intelligence Index

Artificial Analysis intelligence index (Version 2, released Feb 25), Output Tokens Used (~SM input tokens)

Most attractive quadrant

5 vay Geman? A Artificial Anatysis

Pd 70 Grok Jira

i « ‘ re

« Non-Reasoning Models ° Germ.

= * wash

3 GPT-4.1 OeepSeocs \ (Reasoning)

5 55 cota mn v3 0324 tone 3 Claude 3.7

B/S" ee ES — rae

& Prerver

j : Sore coat Reasoning Models

= a e Genwnes

35 Mistral Large

2 (Now 24)

4M Ge 20M 30M 70M 100M 3008

Output Tokens Used In Artificial Analysis Intelligence Index (Log Scale)

A. Artificial Anatysis

slide-006.jpg

OCR text:

OL?

j Reasoning model latency: Reasoning models are slower to provide their response,

making them less suitable for latency sensitive tasks

End-to-End Response Time

supoeciste G camer? wal FT anaes Teh, cE sonny genes Mee Qeomae e eae | ange scudlleageren Reasoning models

NON-EXHAUSTIVE

@ Input processing time @® ‘Thinking’ time (reasoning models) Outputting time

Re i jodel: diy take ge: .

ee ye A) Artificial Anatysis

106.1

po)

60.4

36.9 42.4 43.2 Po

24

eo gi os SS

33 43 AS 8877 ex exa ES bee

~ w” & @ a ae S ” S Se A ¥ ec Se 3 ms Y

mL + 2 5 &€ + & & o m Se Ox fo) > i. £ nN

ge F Sg 8 § SF k& a- gse g8 BS & &e &

SF os ' 8 ¥ é woe ENS Ses So OS & se $£ S SS

r @ ¢ = = é Se FF #F Ce st §8 & gf FF SF

< ¢ x y eo yg £ -

g 6 < é € q 6 F S$ ge we

BR NS 8 A\ Artificial Analysis

slide-007.jpg

OCR text:

@ Cee ee Me Lae

‘ 3 Bere 5 “

| Today's open weights frontier is led by China-based Al labs, namely DeepSeek and Alibaba

Reasoning: Open Weights Language Models, by Country Non-reasoning: Open Weights Language Models, by Country

Artificial Analysis Intelligence Index, leading open weights reasoning models Artificial Analysis Intelligence index, leading open weights non-reasoning models

HB usa Hii China A\ Astiticial Analysis Husa Gi Canada A\ Artificial Analysis

Wi China Ml France

Y

68 wy

eAawo

62 61 60 =

Yo om Yo

mon ma} $3 £2 ween

52st 59 YY $1 Do

48° a7 47 as

OS Pil 5 a

41

a £0 #0 0 38 38

34

DeegSeot Quand Liana3 | DeeoSent Quen? 2B DeapSets Liana i 3 Pn a Mt, = OeenSere Reis DeepSoee Liar d Qwend CLM4.32B Llama} 3) Command Pra MinMa = Gemenal Misra

RiiMay 2358 A278 Neenecron at Reson; AY Death Nemotica ressonng) 7B RL Rt Dear Flash 3 Vo iMer 26) Maverch 235 A278 Inetruct POS. a Yewt.01 27Berwiroct lage?

2075 (Reoscnuny Ultra 25)8 january Qwes 328 Super 498 Lams JOB Now "24>

Reascang 2075: Reason

Feo — er Te As Artificial Anatysis

slide-008.jpg

OCR text:

J Overall cost is a function of both cost per token and tokens per query; we now see a

500X spread in the cost to run Artificial Analysis Intelligence Index

Cost to Run Artificial Analysis Intelligence Index

Cost (USD) ta run all evatuations in the Artificial Analysis intelligence Index (Version 2, released Feb 25, ~5M input tokens) Reasoning models

Input Cost MB Reasoning Cost MM Output Cost NON-EXHAUSTIVE

A) Artificial Analysis

$1952

$1485 $1335

$228

ia $627

Ba

eee be $323, $319 gay

$106 $76 $66 $65 84917 $3 $10 $7 $3

S A S ao e S Y¥ mwa ane 6 & 4 PN Y¥ ” wo S

Ae ° 2 = 6 s £ $ é &. é £ 2 vo > 9

9 ~ dé #23 § VSx *H S82, > oy Sos SF” AS co $

ee EF § SE PSE 8 F Bee BE F Ss

SIS sS EF § g § BF £ § ® : ge* s ge ~S 23 ¥

CSE 68 § Code GES FE we $ &

9° v

SRN At yuts ceepeccrr ties pee un re A» Artificial Anatysis

slide-009.jpg

OCR text:

Overait Costis a function of Doth cost per “oken and tokens per Query: we now see g ; a mo Ps Ps

S500x Spreaq nie eo8o run Att ee at a ae an

Cost to Nun Artficiat Anata ttetgonce ngs, ae ; ae a re ~ oe

POU ene ye BRR ec “SW f riven sre no ae

e ms cay n

Input Cosy Breanne cost Mops cay Aull « axe a ey a me ee

. a — ae

aN: So ;

Sie. mio: one Saeed ; cee a 1 . —

gees $1334 7 ent em —

| m ser Bee kg oo 4 eee ne

ve | cam pie ges Ws ace 7 ee = ae :

| Ed tit “des mo BP es ae 7

" , Awe

slide-010.jpg

OCR text:

[Words Fa lai men err ane _—

: Aiacid S Ripe Aa Pc ae

EN le se a

ae | Arles cs ( Bao OO3 ce Far | WW Windsurf een Fa

ae

Pra Pa ee Sa ail

ea PICO MME iz a @ twiio rena ey

' rae CE ome ens hac '

A |

ee fn a rere 5 an) Rika

[Words Nor am a Scena Mickel LY idslidaiel

ater S

slide-011.jpg

OCR text:

eee Artificial Analysis

ieee ccd

Fc IM I) 2

| HINTNNTNNSY Thsseseses- cssosel =

| | «| ili =. Tinta:

rere a Engineering the future of Al

Slide-Derived Subjects To Review

Subject extraction uses video title, related session titles/descriptions, transcript context, and OCR text when available. OCR is best-effort and should be reviewed against the embedded slide images.