Skip to content
← AI by field
Anyone whose work happens on a computer

AI for remote and knowledge work, right now

Doing paid projects end to end, office documents, research and multi-step work across apps.

Moving: fastUpdated Nov 12, 2026 from 11 sources
Open knowledge work in Spezia Free account. No card.
Where AI is in knowledge work now

Claude Fable 5.2 now completes 41.3% of real freelance projects on the Remote Labor Index, up from 2.1% a year ago.

What models can do now

  • Claude Fable 5.2 completes 41.3% of real freelance projects at professional quality on the Remote Labor Index.[1][5][9][10]
  • Gemini 4 Carbon scores 84.2% on APEX-Agents, a benchmark of long tasks from investment banking, consulting, and law.[2]
  • Claude Fable 5.2 scores 72.9% on OfficeQA Pro, up from 54.1% seven months earlier.[3]
  • Claude Fable 5.2 scores 54.6% on OSWorld 2.1, a benchmark of longer desktop tasks graded at checkpoints.[4]
  • Open-source NorthCUA computer-use models score up to 88 on OS World Verified.[7]

Where they still fall short

  • Claude Fable 5.2 still struggles with animations and industrial product design on the Remote Labor Index.[11]
  • The best model on the Remote Labor Index completes only 41.3% of tasks, leaving most freelance work not automatable.[1][5][9][10]

What is expected next

  • Musk estimates digital AI could add 20-30% to the global economy and double white-collar productivity.[6]
  • Anthropic calls for a verifiable global pause on AI and warns of entry-level white-collar job losses.[8]

How fast it is moving

The record on this field's benchmarks over the last year, from Spezia's model leaderboard.

Remote Labor Index

Real freelance projects that people were paid to do, from design and video to architecture, games and data work. The model gets the same brief and files the freelancer got and has to hand back finished work.

Record: 41.3%, up from 2.1% a year ago (19.9×) (Claude Fable 5.2)

0%15%30%45%60%Claude Fable 5.2Claude Opus 5.5
Jul 25Oct 25Jan 26Apr 26Jul 26Oct 26

the record so far set a record other models

APEX-Agents

Long tasks written by investment banking analysts, management consultants and corporate lawyers from their own work, done across files and apps in a realistic work setup.

Record: 84.2%, up from 21.3% a year ago (4×) (Gemini 4 Carbon)

0%25%50%75%100%Gemini 4 CarbonClaude Sonnet 5.5
Oct 25Jan 26Apr 26Jul 26Oct 26

the record so far set a record other models

  • GDPval-AA1,846, with no earlier result to compare · Claude Opus 5.5
  • OfficeQA Pro72.9%, up from 54.1% 7 months ago · Claude Fable 5.2
  • OSWorld 2.154.6%, up from 43.5% 1 months ago · Claude Fable 5.2
  • METR time horizon12 hours, up from 3.4 hours a year ago (3.5×) · Claude Opus 4.6

The best models for knowledge work now

On this field's benchmarks alone.

  1. 1Claude Fable 5.2
  2. 2Gemini 4 Argon
  3. 3Claude Opus 5.5
  4. 4Claude Sonnet 5.5
  5. 5Claude Fable 5.1
  6. 6Kimi K3.1 Pro

Milestones

What became possible, newest first, each marked by how firm it is. Open one for the words it rests on.

Showndemonstrated with published results, a paper or an independent checkClaimedsaid by the company or people who built it, not yet checked by othersReportedreported by others, not confirmed by the company or a published resultNot publicdone by a model that is not released to the public

  1. Nov 5, 2026ShownNorthCUA
    Open-source NorthCUA computer-use models are released, scoring up to 88 on OS World Verified.
    “on OS World Verified, which is a legit computer control benchmark, we have the Nano coming in at 78.9, the Mini at 85, and the Pro at 88” Open-source NorthCUA computer-use models released, scoring up to 88 on OS World
Every milestone, story and model Pick a model to see what it can do in knowledge work.

Sources

  1. [1]Remote Labor Index: best result 41.3%, up from 2.1% a year ago (19.9×)Leaderboard · 2026-11-09
  2. [2]APEX-Agents: best result 84.2%, up from 21.3% a year ago (4×)Leaderboard · 2026-10-28
  3. [3]OfficeQA Pro: best result 72.9%, up from 54.1% 7 months agoLeaderboard · 2026-11-09
  4. [4]OSWorld 2.1: best result 54.6%, up from 43.5% 1 months agoLeaderboard · 2026-11-09
  5. [5]Fable 5.2 scores 41% on Scale AI and CAIS Remote Labor Index, up from 3.8% a year ago.Spezia news · 2026-11-10
  6. [6]Musk estimates digital AI could add 20-30% to the global economy and double white-collar productivity.Spezia news · 2026-11-10
  7. [7]Open-source NorthCUA computer-use models released, scoring up to 88 on OS World Verified.Spezia news · 2026-11-05
  8. [8]Anthropic calls for a verifiable global pause on AI and warns of entry-level white-collar job losses.Spezia news · 2026-11-02
  9. [9]Scale Labs Remote Labor Index: Claude Fable 5.2 Clears 41.3% of Real Upwork Projects, Double Opus 5.5 | AI WeeklyWeb · 2026-11-09
  10. [10]Fable 5.2 Leads Remote Labor Index with 41.3% Completion RateWeb · 2026-11-09
  11. [11]Anthropic's Fable 5.2 leads Remote Labor Index with 41.3%Web · 2026-11-09

Other fields