Latest
  1. Johns Hopkins Bayview tells 3,182 patients an employee put their health data into a public AI tool
  2. China’s AI compute access expected to double in 2027, with US-designed chips about 65 per cent
  3. UK commits £100m to national AI Biosecurity Lab
  4. Moonshot AI releases Kimi K3.1 Pro with 1-million-token context window
  5. Residents in Newport and Cardiff oppose hyperscale datacentres near homes
  6. Study: Grok 4.7 picked Republicans in 92% of forced midterm answers, GPT-6.1 Sol picked Democrats in 97%
  7. Attackers exploiting the PHP flaws hand the break-in to AI agents, Unit 42 says
  8. Elvis Presley’s rightsholders sue Anthropic over training on 274 songs

Latest / Finance

ResearchFinanceFrontierUSConfirmed

Bloomberg describes BloombergGPT, a 50-billion-parameter model trained on financial data

See how this story connects

Source: arXiv

Bloomberg researchers posted a paper on arXiv on 30 March 2023 describing BloombergGPT, a 50-billion-parameter language model. It was trained on 363 billion tokens from Bloomberg’s own financial data sources plus 345 billion tokens of general-purpose text. The authors report that it beats existing open models on financial tasks by significant margins while holding its scores on general benchmarks.

Why it matters

A financial data company trained its own large language model on its proprietary data and published the method and results for banks deciding whether to build or buy models.

When
30 Mar 2023
ID
WTA-N-2023-0050
Who
Bloomberg

Summary by wheresthe.ai, from arXiv. Checked against the source on 9 Oct 2026.

Read the original at arXiv

Source copy: download the page as saved on 18 Oct 2026 (PDF). Also held by the Internet Archive.

Next stop on this Storyline: What happened nextAnt Group unveils a financial large language model and two assistants built on it

Cite this story

WTA-N-2023-0050: “Bloomberg describes BloombergGPT, a 50-billion-parameter model trained on financial data.” wheresthe.ai, 30 Mar 2023. https://wheresthe.ai/d/bloomberg-bloomberggpt-paper-2023-03/ (source: arXiv).

Spotted an error? Email [email protected].