Vals AI, 2026

Vals Finance Index

A composite score for agentic finance work. It averages Finance Agent v2.1 and EMB v1.1 for securities, adds Tax Agent Bench for tax, and weights the two by each sub-sector's share of US GDP (1.6 vs 0.5).

Best reported

66.3

Gemini 4 CarbonGoogle

index accuracy %

Official board, 29 Oct 2026

About this benchmark

Size
Composite of 3 benchmarks
Public split
none (inherits the public samples of its components)
Grading
Index = (1.6 x avg(FAB v2.1, EMB v1.1) + 0.5 x Tax Agent Bench) / 2.1. A model gets a score only if it has results on all three.
Licence
Not stated
Submission
Not stated
Products allowed
Not stated
Latest version
v1 (released 2026-10-21)
What it does not cover
Nothing beyond its three components: no trading, advice or market-data tasks.

Results

10 models
Leaderboard

Each score shows who reported it: the benchmark's authors (quoted from their paper), the benchmark's official board, or Moneda (run with the benchmark's own grader and rules). We do not copy other aggregators.

Best value: Wonder Full, 64.5 at $3.14 per question (cheapest within 5% of the top score)

Vals Finance Index: score vs cost

Score against cost per test. Up and to the left is better.

Updated 29 Oct 2026

Y: index accuracy %X: Cost per test (USD)Showing 10 models
Vals Finance Index: score vs cost10 models. Frontier: DeepSeek V4.1 Flash 1017, Wonder Full, Gemini 4 Carbon.60626466$0$2.5$5$7.5$10$12.5EfficientInefficient↑ Best performance← Lowest costCost per test (USD)Gemini 4 Carbon: 66.3 index accuracy %, $6.46 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)Gemini 4 CarbonClaude Opus 5.5: 65.6 index accuracy %, $11.67 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)Claude Opus 5.5Gemini 4 Argon: 64.9 index accuracy %, $6.53 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)Gemini 4 ArgonWonder Full: 64.5 index accuracy %, $3.14 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)Wonder FullClaude Sonnet 5.5: 64.5 index accuracy %, $8.33 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)Claude Sonnet 5.5Qwen 3.8 Max 0902: 61.7 index accuracy %, $3.83 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)Qwen 3.8 Max 0902Wonder: 61.2 index accuracy %, $3.37 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)WonderDeepSeek V4.1 Flash 1017: 60.9 index accuracy %, $0.42 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)DeepSeek V4.1 Flash 1017GLM 5.3: 60.5 index accuracy %, $2.21 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026) GLM 5.3Gemini 3.8 Flash: 60.5 index accuracy %, $3.7 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)Gemini 3.8 FlashVals Finance Index: score vs cost10 models. Frontier: DeepSeek V4.1 Flash 1017, Wonder Full, Gemini 4 Carbon.60626466$0.5$1$2$5$10EfficientInefficient↑ Best performance← Lowest costCost per test (USD)Gemini 4 Carbon: 66.3 index accuracy %, $6.46 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)Gemini 4 CarbonClaude Opus 5.5: 65.6 index accuracy %, $11.67 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)Claude Opus 5.5Gemini 4 Argon: 64.9 index accuracy %, $6.53 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)Gemini 4 ArgonWonder Full: 64.5 index accuracy %, $3.14 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)Wonder FullClaude Sonnet 5.5: 64.5 index accuracy %, $8.33 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)Claude Sonnet 5.5Qwen 3.8 Max 0902: 61.7 index accuracy %, $3.83 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)Qwen 3.8 Max 0902Wonder: 61.2 index accuracy %, $3.37 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)WonderDeepSeek V4.1 Flash 1017: 60.9 index accuracy %, $0.42 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)DeepSeek V4.1 Flash 1017GLM 5.3: 60.5 index accuracy %, $2.21 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026) GLM 5.3Gemini 3.8 Flash: 60.5 index accuracy %, $3.7 per test (Vals.ai Finance Index leaderboard, Cost / Test, 29 Oct 2026)Gemini 3.8 Flash

Vals Finance Index v1 (released 2026-10-21)

index accuracy %, higher is better. 10 models.

Updated 29 Oct 2026

  1. Gemini 4 Carbon66.3
  2. Claude Opus 5.565.6
  3. Gemini 4 Argon64.9
  4. Wonder Full64.5
  5. Claude Sonnet 5.564.5
  6. Qwen 3.8 Max 090261.7
  7. Wonder61.2
  8. DeepSeek V4.1 Flash 101760.9
  9. GLM 5.360.5
  10. Gemini 3.8 Flash60.5

Vals Finance Index: cost

cost per test item (USD), lower is better. 10 models.

Updated 29 Oct 2026

  1. DeepSeek V4.1 Flash 10170.420
  2. GLM 5.32.2
  3. Wonder Full3.1
  4. Wonder3.4
  5. Gemini 3.8 Flash3.7
  6. Qwen 3.8 Max 09023.8
  7. Gemini 4 Carbon6.5
  8. Gemini 4 Argon6.5
  9. Claude Sonnet 5.58.3
  10. Claude Opus 5.511.7

Vals Finance Index v1 (released 2026-10-21)

index accuracy %, higher is better. 10 models.

Updated 29 Oct 2026

  1. 1Gemini 4 Carbon66.3
  2. 2Claude Opus 5.565.6
  3. 3Gemini 4 Argon64.9
  4. 4Wonder Full64.5
  5. 5Claude Sonnet 5.564.5
  6. 6Qwen 3.8 Max 090261.7
  7. 7Wonder61.2
  8. 8DeepSeek V4.1 Flash 101760.9
  9. 9GLM 5.360.5
  10. 10Gemini 3.8 Flash60.5
Vals Finance Index leaderboard (index accuracy %, higher is better)
1Gemini 4 CarbonGoogle66.3DefaultVals.ai leaderboard
Official board, 29 Oct 2026
2Claude Opus 5.5Anthropic65.6DefaultVals.ai leaderboard
Official board, 29 Oct 2026
3Gemini 4 ArgonGoogle64.9DefaultVals.ai leaderboard
Official board, 29 Oct 2026
4Wonder FullNot stated64.5DefaultVals.ai leaderboard
Official board, 29 Oct 2026
5Claude Sonnet 5.5Anthropic64.5DefaultVals.ai leaderboard
Official board, 29 Oct 2026
6Qwen 3.8 Max 0902Qwen61.7DefaultVals.ai leaderboard
Official board, 29 Oct 2026
7WonderNot stated61.2DefaultVals.ai leaderboard
Official board, 29 Oct 2026
8DeepSeek V4.1 Flash 1017DeepSeek60.9DefaultVals.ai leaderboard
Official board, 29 Oct 2026
9GLM 5.3Zhipu AI60.5DefaultVals.ai leaderboard
Official board, 29 Oct 2026
10Gemini 3.8 FlashGoogle60.5DefaultVals.ai leaderboard
Official board, 29 Oct 2026

Last checked 1 Nov 2026. Something out of date or wrong? Send a correction.