Vals AI Financial Coaching Bench (Financial Coaching Bench)
We show this table for reference; we do not rank on it.
Can AI coach a household through a real financial problem?
Evaluation results published by Vals AI. Published leaderboard aggregates and evaluation settings only. Dataset contents are not included.
Checklist items met on Financial Coaching Bench — November 13, 2026
We mirror the published checklist items met view for Financial Coaching Bench. Claude Opus 5.5 leads the public snapshot at 76.16%, followed by Muse Spark 1.2 (75.74%) and Claude Fable 5.1 (75.71%). We do not use these results to rank models overall.
Claude Opus 5.5
Anthropic
Muse Spark 1.2
Meta
xhigh reasoning
Claude Fable 5.1
Anthropic
28 modelsAgenticCurrentDisplay onlyUpdated November 13, 2026
Checklist items met table (28 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Claude Opus 5.5Anthropic | 76.16% | Not reported | Closed |
| 2 | Muse Spark 1.2Metaxhigh reasoning | 75.74% | Not reported | Closed |
| 3 | Claude Fable 5.1Anthropic | 75.71% | Not reported | Closed |
| 4 | Qwen3.8 MaxAlibaba | 75.58% | Not reported | Open |
| 5 | MiMo-V2.6-ProXiaomi | 75.37% | Not reported | Open |
| 6 | Muse Wonder FullMetamax reasoning | 74.96% | Not reported | Closed |
| 7 | Claude Sonnet 5.5Anthropic | 74.94% | Not reported | Closed |
| 8 | GPT-5.6 SolOpenAImax reasoning | 74.89% | Not reported | Closed |
| 9 | Kimi K3Moonshot AImax reasoning | 74.53% | Not reported | Open |
| 10 | Grok 4.7xAIxhigh reasoning | 74.07% | Not reported | Closed |
| 11 | GLM-5.3Z.AImax reasoning | 74.06% | Not reported | Open |
| 12 | Muse WonderMetamax reasoning | 73.78% | Not reported | Closed |
| 13 | Gemini 4 ArgonGooglehigh reasoning | 73.13% | Not reported | Closed |
| 14 | DeepSeek V4.1 Flash 1017DeepSeekhigh reasoning | 72.83% | Not reported | Open |
| 15 | Muse Spark 1.3Metaxhigh reasoning | 70.99% | Not reported | Closed |
| 16 | GPT-5.6 TerraOpenAImax reasoning | 70.98% | Not reported | Closed |
| 17 | Muse Spark 1.3 MaxMetamax reasoning | 70.29% | Not reported | Unknown |
| 18 | GPT-5.6 LunaOpenAImax reasoning | 69.29% | Not reported | Closed |
| 19 | MiniMax M3MiniMax | 69.05% | Not reported | Open |
| 20 | GPT-6.1 SolOpenAImax reasoning | 68.55% | Not reported | Closed |
| 21 | GPT-6 AstraOpenAImax reasoning | 67.53% | Not reported | Closed |
| 22 | Gemini 3.1 Pro PreviewGooglehigh reasoning | 66.73% | Not reported | Unknown |
| 23 | Gemini 3.8 FlashGooglehigh reasoning | 66.27% | Not reported | Closed |
| 24 | Gemini 3.6 FlashGooglehigh reasoning | 65.76% | Not reported | Closed |
| 25 | GPT-6 LunaOpenAImax reasoning | 65.24% | Not reported | Closed |
| 26 | Gemini 3.5 FlashGooglehigh reasoning | 62.96% | Not reported | Closed |
| 27 | GPT-6 SolOpenAImax reasoning | 62.61% | Not reported | Closed |
| 28 | Gemini 3.5 Flash-LiteGooglehigh reasoning | 57.47% | Not reported | Closed |
How to read this leaderboard
Operator receipt: 28 sourced rows are currently displayable on this page; the leading published row is Claude Opus 5.5 at 76.16%.
Honest limit: The test set is private except two example scenarios. Borrowing makes up 28% of scenarios and digital-first behaviors 8%. Claude Sonnet 4.6 plays every client, and Vals reports that longer answers score higher within a model, so the score partly rewards length. Models were run through their APIs, which approximates but does not reproduce what a chatbot user sees.
How Financial Coaching Bench is shown here
We mirror Vals AI's Financial Coaching Bench, published November 13, 2026 with support from NextLadder Ventures and advice from the Financial Health Network. It asks whether a model can coach a household through a real money problem, such as a collections notice, a rent increase or a defaulted loan.
Each of the 256 scored scenarios starts with a client's opening message and a hidden fact sheet of income, debts, deadlines and savings. The model plays the coach: it can ask questions and search the web while a simulated client answers from the fact sheet. Practitioners wrote a checklist of what a good coach recommends, warns about or gets right, and the score is the share of checklist items met. Scenarios cover seven areas led by borrowing (28%). The results stay display only and do not enter model rankings.
Snapshot
The published Financial Coaching Bench snapshot places Claude Opus 5.5 first at 76.16%. The third row is 0.45 points behind. The broader top-10 range is 2.09 points, so many of the published results sit in a relatively narrow band.
28 models have been evaluated on Financial Coaching Bench. The benchmark falls in the Agentic category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. Financial Coaching Bench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About Financial Coaching Bench
Year
2026
Tasks
256 simulated coaching conversations across 7 financial areas
Format
Share of practitioner checklist items met
Difficulty
Household financial coaching with follow-up questions
Vals published Financial Coaching Bench on November 13, 2026. Each of 256 scored scenarios gives the model a client's opening message about a money problem, such as a collections notice, a rent increase or a defaulted loan. A simulated client answers the coach's questions from a hidden fact sheet of income, debts, deadlines, savings and household. The coach can search the web and ends by delivering a plan. Practitioners wrote a checklist of what a good coach recommends, warns about or gets right, and the score is the share of checklist items met. BenchLM mirrors the published aggregates as display-only external evidence, outside overall and category rankings.
Freshness and provenance
Version
Financial Coaching Bench 2026
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 Financial Coaching Bench measure?
Can AI coach a household through a real financial problem?
Which model leads the published Financial Coaching Bench snapshot?
Claude Opus 5.5 currently leads the published Financial Coaching Bench snapshot with 76.16% checklist items met. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on Financial Coaching Bench?
The November 13, 2026 snapshot contains 28 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.