FinanceArena — FinanceQA Assumption-Based (FinanceArena)
We show this table for reference; we do not rank on it.
An AfterQuery benchmark of open-ended financial analysis that requires models to read financial data, make assumptions, and return exact answers.
No longer maintained · AfterQuery took this leaderboard down in October 2026, so it no longer updates. The table below is the last version we captured (October 12, 2026 snapshot) and stays here for reference.
Exact-match accuracy on FinanceQA Assumption-Based — October 12, 2026 snapshot
We mirror the published exact-match accuracy view for FinanceQA Assumption-Based. o3 leads the public snapshot at 54.1%, followed by Grok 4 (49.3%) and o4-mini (high) (48.6%). We do not use these results to rank models overall.
o3
OpenAI
Grok 4
xAI
o4-mini (high)
OpenAI
19 modelsKnowledgeCurrentDisplay onlyUpdated October 12, 2026 snapshot
Exact-match accuracy table (19 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | o3OpenAI | 54.1% | Not reported | Closed |
| 2 | Grok 4xAI | 49.3% | Not reported | Closed |
| 3 | o4-mini (high)OpenAI | 48.6% | Not reported | Closed |
| 4 | Gemini 2.5 ProGoogle | 45.3% | Not reported | Closed |
| 5 | Llama 4 MaverickMeta | 44.6% | Not reported | Open |
| 6 | Claude Opus 4Anthropic | 44.6% | Not reported | Unknown |
| 7 | Grok 3 [Beta]xAI | 44.6% | Not reported | Closed |
| 8 | Claude Sonnet 4Anthropic | 43.9% | Not reported | Unknown |
| 9 | Phi-4-reasoning-plusMicrosoft | 43.2% | Not reported | Open |
| 10 | DeepSeek-R1DeepSeek | 42.9% | Not reported | Open |
| 11 | QwQ-32BAlibaba | 42.6% | Not reported | Open |
| 12 | GPT-4.1 miniOpenAI | 41.9% | Not reported | Closed |
| 13 | Llama 3.1 Nemotron UltraNVIDIA | 41.9% | Not reported | Unknown |
| 14 | Qwen3 30B A3BAlibaba | 37.2% | Not reported | Unknown |
| 15 | Kimi K2Moonshot AI | 33.8% | Not reported | Closed |
| 16 | Gemini 2.5 FlashGoogle | 32.4% | Not reported | Closed |
| 17 | Magistral MediumMistral | 31.8% | Not reported | Closed |
| 18 | Command ACohere | 27.7% | Not reported | Unknown |
| 19 | Nova ProAmazon | 20.3% | Not reported | Closed |
How to read this leaderboard
Operator receipt: 19 sourced rows are currently displayable on this page; the leading published row is o3 at 54.1%.
Honest limit: AfterQuery's page describes three FinanceQA task families, but its embedded leaderboard payload labels this table Assumption-Based without publishing a task count or row-level protocol fields. Read the table as an exact source mirror, not as a normalized finance ranking.
How we show FinanceArena
We mirror AfterQuery's FinanceQA, Assumption-Based table captured on October 12, 2026 snapshot. FinanceArena grades professional financial-analysis answers with exact-match accuracy and publishes 19 model rows in this view.
AfterQuery labels the table “FinanceQA, Assumption-Based,” while the embedded leaderboard payload does not expose a task count or per-row run settings. We preserve that source label and keep the scores display-only instead of merging them with other finance evaluations.
Snapshot
The published FinanceArena snapshot places o3 first at 54.1%. The third row is 5.5 points behind. The broader top-10 range is 11.2 points, so the table still separates the published systems.
19 models have been evaluated on FinanceArena. The benchmark falls in the Knowledge category. FinanceArena is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About FinanceArena
Year
2025
Tasks
Professional financial-analysis questions
Format
Open-ended financial QA with exact-match grading
Difficulty
Professional finance reasoning
The mirrored AfterQuery table is labeled FinanceQA, Assumption-Based and reports exact-match accuracy. We keep it display-only because the embedded leaderboard rows do not expose the task count or per-row run configuration needed to merge them with other finance evaluations.
Freshness and provenance
Version
FinanceArena 2025
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does FinanceArena measure?
An AfterQuery benchmark of open-ended financial analysis that requires models to read financial data, make assumptions, and return exact answers.
Which model leads the published FinanceArena snapshot?
o3 currently leads the published FinanceArena snapshot with 54.1% exact-match accuracy. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on FinanceArena?
The October 12, 2026 snapshot snapshot contains 19 AI models.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.