BenchLeader
Menu
LeaderboardsModelsProvidersBenchmarksCompareReleases1BlogMethodology

Finance Agent v2

Core financial-analyst tasks with tools. Run by Vals AI.

As of 19 Sept 2026, Gemini 3.8 Flash leads Finance Agent v2 on BenchLeader with 61.4%, ahead of Muse Spark 1.2 at 60.6%, across 58 model configurations with a published result.

Published by
Vals AI
Category
Agents & tools
Index weight
Reference only
Models
58
Data as of
19 Sept 2026

Vals AI (vals.ai).

What the test looks like

Tasks a junior financial analyst does: pulling figures from filings, computing ratios, reconciling numbers, with tools available.

How it is scored

Accuracy across tasks, run by Vals AI.

What to keep in mind

Overlaps with CorpFin; this version is agentic and tool-using.

Top 15
  1. 1Gemini 3.8 Flash (high)61.4%
  2. 2Muse Spark 1.2 (xhigh)60.6%
  3. 3Muse Spark 1.3 (max)60.0%
  4. 4Gemini 3.7 Flash (high)59.0%
  5. 5Muse Spark 1.3 (xhigh)58.9%
  6. 6Claude Fable 5.158.9%
  7. 7Claude Opus 558.6%
  8. 8Gemini 3.5 Flash (high)57.9%
  9. 9GLM-5.3-Flash (max)57.9%
  10. 10Muse Spark 1.1 (xhigh)57.2%
  11. 11Claude Fable 556.3%
  12. 12Gemini 3.6 Flash (high)56.3%
  13. 13GLM-5.3 (max)55.8%
  14. 14GPT-5.6 Luna (max)55.0%
  15. 15GPT-5.6 Terra (max)54.4%
58 of 58
#
1Gemini 3.8 FlashhighGoogle61.4%64.52026-09-15
2Muse Spark 1.2xhighMeta60.6%64.12026-09-15
3Muse Spark 1.3maxMeta60.0%69.32026-09-15
4Gemini 3.7 FlashhighGoogle59.0%64.62026-09-15
5Muse Spark 1.3xhighMeta58.9%68.02026-09-15
6Claude Fable 5.1Anthropic58.9%64.52026-09-15
7Claude Opus 5Anthropic58.6%66.72026-09-15
8Gemini 3.5 FlashhighGoogle57.9%63.62026-09-15
9GLM-5.3-FlashmaxZhipu AIopen ↗57.9%55.32026-09-15
10Muse Spark 1.1xhighMeta57.2%61.82026-09-15
11Claude Fable 5Anthropic56.3%68.32026-09-15
12Gemini 3.6 FlashhighGoogle56.3%61.72026-09-15
13GLM-5.3maxZhipu AIopen ↗55.8%65.82026-09-15
14GPT-5.6 LunamaxOpenAI55.0%60.22026-09-15
15GPT-5.6 TerramaxOpenAI54.4%65.02026-09-15
16Kimi K3Moonshot AIopen ↗54.4%64.22026-09-15
17Claude Opus 4.8Anthropic53.9%61.92026-09-15
18Claude Sonnet 5Anthropic53.9%57.22026-09-15
19GPT-5.6 SolmaxOpenAI53.8%68.82026-09-15
20Grok 4.6highxAI53.7%64.52026-09-15
21GPT-6 AstramaxOpenAI53.5%71.82026-09-15
22DeepSeek V4.1 FlashhighDeepSeekopen ↗53.5%–2026-09-15
23GPT-5.5xhighOpenAI51.8%67.62026-09-15
24Claude Opus 4.7Anthropic51.5%64.52026-09-15
25Claude Sonnet 4.6Anthropic51.0%59.02026-09-15
26Qwen3 8maxAlibaba50.6%66.52026-09-15
27DeepSeek V4 Pro 0813maxDeepSeekopen ↗50.4%56.02026-09-15
28GLM-5.2Zhipu AIopen ↗49.7%52.12026-09-15
29DeepSeek V4 Flash 0731highDeepSeekopen ↗49.5%55.02026-09-15
30Qwen3.8 27BxhighAlibabaopen ↗48.5%59.22026-09-15
31Grok 4.5highxAI48.4%61.02026-09-15
32MiniMax-M3MiniMaxopen ↗48.3%57.52026-09-15
33Qwen3 7maxAlibaba47.8%61.02026-09-15
34Gemini 3.5 Flash LitehighGoogle47.4%44.32026-09-15
35GPT-5.4 minixhighOpenAI45.4%55.72026-09-15
36Kimi K2.6Moonshot AIopen ↗44.9%60.52026-09-15
37GLM-5.1Zhipu AIopen ↗44.8%56.92026-09-15
38DeepSeek V4 PromaxDeepSeekopen ↗44.1%64.22026-09-15
39Gemini 3.1 ProhighGoogle43.0%60.72026-09-15
40Gemini 3 FlashhighGoogle42.5%58.72026-09-15
41MiMo-V2.5-ProXiaomiopen ↗41.5%59.22026-09-15
42Inkling SmallThinking Machinesopen ↗41.3%56.32026-09-15
43Qwen3.6 PlusAlibabaopen40.9%58.52026-09-15
44Qwen3.7 PlusAlibaba38.2%59.92026-09-15
45GPT-5.4 nanohighOpenAI38.2%49.12026-09-15
46Grok 4.3highxAI37.7%58.22026-09-15
47Nemotron 3 Ultra 550B A55BNVIDIAopen ↗37.7%48.12026-09-15
48Kimi K2.5thinkingMoonshot AIopen35.8%58.52026-09-15
49Mistral Medium 3.5highMistral AIopen32.1%–2026-09-15
50Claude Haiku 4.5thinkingAnthropic31.0%47.52026-09-15
51Ling 3.0 FlashAnt Groupopen ↗30.3%53.62026-09-15
52Gemini 3.1 Flash LitehighGoogle30.0%49.12026-09-15
53Grok 4.20thinkingxAI28.5%59.92026-09-15
54MiniMax-M2.7MiniMaxopen ↗27.9%56.12026-09-15
55Laguna M.1Poolsideopen25.0%37.72026-09-15
56Mercury 2.5highInception18.6%–2026-09-15
57Nemotron Lightning 3.5 30B A3bNVIDIA18.5%–2026-09-15
58Laguna XS.2Poolside15.6%36.52026-09-15

Cite as: BenchLeader, “Finance Agent v2 leaderboard”, https://www.benchleader.com/benchmarks/vals_finance_agent, data as of 19 Sept 2026.

Finance Agent v2: questions

What does Finance Agent v2 measure?
Tasks a junior financial analyst does: pulling figures from filings, computing ratios, reconciling numbers, with tools available. Scores are reported in percent of tasks solved; higher is better.
Which AI model leads Finance Agent v2?
Gemini 3.8 Flash leads Finance Agent v2 with 61.4% as of 19 Sept 2026, ahead of Muse Spark 1.2 at 60.6%.
How many models have Finance Agent v2 results?
58 model configurations have a Finance Agent v2 result on BenchLeader, all taken from Vals AI.
Who runs Finance Agent v2 and how often is it updated?
Finance Agent v2 is published by Vals AI. BenchLeader re-reads the published results every morning and records the date each result was published.
Does Finance Agent v2 count toward the BenchLeader Index?
No. Finance Agent v2 is shown for reference but left out of the composite index.