Skip to content
← AI by field
Analysts, bankers, investors and finance teams

AI for finance, right now

Analyzing filings and markets, building models and spreadsheets, accounting and audit work.

Moving: fastUpdated Nov 12, 2026 from 9 sources
Open finance in Spezia Free account. No card.
Where AI is in finance now

AI models now handle real financial analysis tasks, with Claude Fable 5.2 leading finance benchmarks.

What models can do now

  • Claude Fable 5.2 leads finance benchmarks, followed by Claude Opus 5.5 and Claude Sonnet 5.5.[5]
  • Gemini 3.5 Flash achieved 57.9% on Finance Agent v2, up from 31.0% a year ago.[1]
  • Gemini 4 Carbon achieved 84.2% on APEX-Agents, up from 21.3% a year ago.[2]
  • Claude Fable 5.2 achieved 72.9% on OfficeQA Pro, up from 54.1% seven months ago.[3]
  • Claude Opus 5.5 achieved 1,846 on GDPval-AA, with no earlier result to compare.[4]
  • Northern Trust and Microsoft launched an AI agent that cuts client onboarding data preparation time by 83%, from two and a half weeks to three days.[9]

Where they still fall short

  • AI auditors detect reward hacking well but flag harmless logs at higher rates, according to an Anthropic Institute researcher.[8]

What is expected next

  • OpenAI's chief economist projects AI could add one percentage point to annual productivity growth within 18-24 months.[7]
  • J.P. Morgan Private Bank forecasts AI will lift U.S. labor productivity by 50bps to 2.2% annually.[6]

How fast it is moving

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

Finance Agent v2

Real financial analysis work: finding figures in company filings and documents, multi-step calculations and written analysis.

Record: 57.9%, up from 31.0% a year ago (1.9×) (Gemini 3.5 Flash)

0%15%30%45%60%GPT-5.5Gemini 3.5 Flash
Jan 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

  • OfficeQA Pro72.9%, up from 54.1% 7 months ago · Claude Fable 5.2
  • GDPval-AA1,846, with no earlier result to compare · Claude Opus 5.5

The best models for finance now

On this field's benchmarks alone.

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

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 9, 2026ShownMicrosoft Azure AI agent
    Northern Trust and Microsoft launched an AI agent that converts historical client transactions into standardized reporting data, cutting preparation time by 83%.
    “Northern Trust’s production measurements showed preparation time fell 83%, from about two and a half weeks to three days” Northern Trust, Microsoft Launch AI Agent to Speed Client Onboarding｜Mark Radar
  2. Nov 9, 2026ShownClaude Fable 5.2
    Anthropic's Claude Fable 5.2 scored 81.2% on the DualEntry Accounting AI Benchmark, a 3.9-point decline from Fable 5.1.
    “Anthropic's newly released Claude Fable 5.2 scored 81.2% on the DualEntry Accounting AI Benchmark” Anthropic's Claude Fable 5.2 Scores Below Fable 5.1 on Accounting Work, DualEntr
  3. Nov 6, 2026Shown
    DeepFinance benchmark tasks test whether an AI agent can complete due diligence on a $100 million deal, with 136 rubric criteria.
    “The task’s rubric has 136 criteria.” DeepFinance: Can AI Diligence a $100MM Deal?
  4. Aug 25, 2026ShownGemini Enterprise for Financial Services
    Google Cloud launched Gemini Enterprise for Financial Services with a Financial Research agent supported by more than 50 foundational skills.
    “Announced on August 25, 2026, the solution is initially available in preview for capital markets and corporate banking” Gemini Enterprise for Financial Services | Prolifics
Every milestone, story and model Pick a model to see what it can do in finance.

Sources

  1. [1]Finance Agent v2: best result 57.9%, up from 31.0% a year ago (1.9×)Leaderboard · 2026-05-19
  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]GDPval-AA: best result 1,846, with no earlier result to compareLeaderboard · 2026-09-22
  5. [5]Best models on finance benchmarksLeaderboard · 2026-11-12
  6. [6]J.P. Morgan Private Bank forecasts AI will lift U.S. labor productivity by 50bps to 2.2% annually.Spezia news · 2026-11-09
  7. [7]OpenAI's chief economist says AI could add a percentage point to annual productivity growth within 18-24 months.Spezia news · 2026-11-09
  8. [8]Anthropic Institute researcher finds AI auditors detect reward hacking well but flag harmless logs at higher rates.Spezia news · 2026-11-08
  9. [9]Northern Trust, Microsoft Launch AI Agent to Speed Client Onboarding｜Mark RadarWeb · 2026-11-09

Other fields